2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/statfs.h>
37 #include <linux/compat.h>
38 #include <linux/bit_spinlock.h>
39 #include <linux/security.h>
40 #include <linux/xattr.h>
41 #include <linux/vmalloc.h>
42 #include <linux/slab.h>
43 #include <linux/blkdev.h>
44 #include <linux/uuid.h>
45 #include <linux/btrfs.h>
46 #include <linux/uaccess.h>
49 #include "transaction.h"
50 #include "btrfs_inode.h"
51 #include "print-tree.h"
54 #include "inode-map.h"
56 #include "rcu-string.h"
58 #include "dev-replace.h"
64 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
65 * structures are incorrect, as the timespec structure from userspace
66 * is 4 bytes too small. We define these alternatives here to teach
67 * the kernel about the 32-bit struct packing.
69 struct btrfs_ioctl_timespec_32 {
72 } __attribute__ ((__packed__));
74 struct btrfs_ioctl_received_subvol_args_32 {
75 char uuid[BTRFS_UUID_SIZE]; /* in */
76 __u64 stransid; /* in */
77 __u64 rtransid; /* out */
78 struct btrfs_ioctl_timespec_32 stime; /* in */
79 struct btrfs_ioctl_timespec_32 rtime; /* out */
81 __u64 reserved[16]; /* in */
82 } __attribute__ ((__packed__));
84 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
85 struct btrfs_ioctl_received_subvol_args_32)
89 static int btrfs_clone(struct inode *src, struct inode *inode,
90 u64 off, u64 olen, u64 olen_aligned, u64 destoff);
92 /* Mask out flags that are inappropriate for the given type of inode. */
93 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
97 else if (S_ISREG(mode))
98 return flags & ~FS_DIRSYNC_FL;
100 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
104 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
106 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
108 unsigned int iflags = 0;
110 if (flags & BTRFS_INODE_SYNC)
111 iflags |= FS_SYNC_FL;
112 if (flags & BTRFS_INODE_IMMUTABLE)
113 iflags |= FS_IMMUTABLE_FL;
114 if (flags & BTRFS_INODE_APPEND)
115 iflags |= FS_APPEND_FL;
116 if (flags & BTRFS_INODE_NODUMP)
117 iflags |= FS_NODUMP_FL;
118 if (flags & BTRFS_INODE_NOATIME)
119 iflags |= FS_NOATIME_FL;
120 if (flags & BTRFS_INODE_DIRSYNC)
121 iflags |= FS_DIRSYNC_FL;
122 if (flags & BTRFS_INODE_NODATACOW)
123 iflags |= FS_NOCOW_FL;
125 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
126 iflags |= FS_COMPR_FL;
127 else if (flags & BTRFS_INODE_NOCOMPRESS)
128 iflags |= FS_NOCOMP_FL;
134 * Update inode->i_flags based on the btrfs internal flags.
136 void btrfs_update_iflags(struct inode *inode)
138 struct btrfs_inode *ip = BTRFS_I(inode);
139 unsigned int new_fl = 0;
141 if (ip->flags & BTRFS_INODE_SYNC)
143 if (ip->flags & BTRFS_INODE_IMMUTABLE)
144 new_fl |= S_IMMUTABLE;
145 if (ip->flags & BTRFS_INODE_APPEND)
147 if (ip->flags & BTRFS_INODE_NOATIME)
149 if (ip->flags & BTRFS_INODE_DIRSYNC)
152 set_mask_bits(&inode->i_flags,
153 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
158 * Inherit flags from the parent inode.
160 * Currently only the compression flags and the cow flags are inherited.
162 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
169 flags = BTRFS_I(dir)->flags;
171 if (flags & BTRFS_INODE_NOCOMPRESS) {
172 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
173 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
174 } else if (flags & BTRFS_INODE_COMPRESS) {
175 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
176 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
179 if (flags & BTRFS_INODE_NODATACOW) {
180 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
181 if (S_ISREG(inode->i_mode))
182 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
185 btrfs_update_iflags(inode);
188 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
190 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
191 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
193 if (copy_to_user(arg, &flags, sizeof(flags)))
198 static int check_flags(unsigned int flags)
200 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
201 FS_NOATIME_FL | FS_NODUMP_FL | \
202 FS_SYNC_FL | FS_DIRSYNC_FL | \
203 FS_NOCOMP_FL | FS_COMPR_FL |
207 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
213 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
215 struct inode *inode = file_inode(file);
216 struct btrfs_inode *ip = BTRFS_I(inode);
217 struct btrfs_root *root = ip->root;
218 struct btrfs_trans_handle *trans;
219 unsigned int flags, oldflags;
222 unsigned int i_oldflags;
225 if (!inode_owner_or_capable(inode))
228 if (btrfs_root_readonly(root))
231 if (copy_from_user(&flags, arg, sizeof(flags)))
234 ret = check_flags(flags);
238 ret = mnt_want_write_file(file);
242 mutex_lock(&inode->i_mutex);
244 ip_oldflags = ip->flags;
245 i_oldflags = inode->i_flags;
246 mode = inode->i_mode;
248 flags = btrfs_mask_flags(inode->i_mode, flags);
249 oldflags = btrfs_flags_to_ioctl(ip->flags);
250 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
251 if (!capable(CAP_LINUX_IMMUTABLE)) {
257 if (flags & FS_SYNC_FL)
258 ip->flags |= BTRFS_INODE_SYNC;
260 ip->flags &= ~BTRFS_INODE_SYNC;
261 if (flags & FS_IMMUTABLE_FL)
262 ip->flags |= BTRFS_INODE_IMMUTABLE;
264 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
265 if (flags & FS_APPEND_FL)
266 ip->flags |= BTRFS_INODE_APPEND;
268 ip->flags &= ~BTRFS_INODE_APPEND;
269 if (flags & FS_NODUMP_FL)
270 ip->flags |= BTRFS_INODE_NODUMP;
272 ip->flags &= ~BTRFS_INODE_NODUMP;
273 if (flags & FS_NOATIME_FL)
274 ip->flags |= BTRFS_INODE_NOATIME;
276 ip->flags &= ~BTRFS_INODE_NOATIME;
277 if (flags & FS_DIRSYNC_FL)
278 ip->flags |= BTRFS_INODE_DIRSYNC;
280 ip->flags &= ~BTRFS_INODE_DIRSYNC;
281 if (flags & FS_NOCOW_FL) {
284 * It's safe to turn csums off here, no extents exist.
285 * Otherwise we want the flag to reflect the real COW
286 * status of the file and will not set it.
288 if (inode->i_size == 0)
289 ip->flags |= BTRFS_INODE_NODATACOW
290 | BTRFS_INODE_NODATASUM;
292 ip->flags |= BTRFS_INODE_NODATACOW;
296 * Revert back under same assuptions as above
299 if (inode->i_size == 0)
300 ip->flags &= ~(BTRFS_INODE_NODATACOW
301 | BTRFS_INODE_NODATASUM);
303 ip->flags &= ~BTRFS_INODE_NODATACOW;
308 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
309 * flag may be changed automatically if compression code won't make
312 if (flags & FS_NOCOMP_FL) {
313 ip->flags &= ~BTRFS_INODE_COMPRESS;
314 ip->flags |= BTRFS_INODE_NOCOMPRESS;
316 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
317 if (ret && ret != -ENODATA)
319 } else if (flags & FS_COMPR_FL) {
322 ip->flags |= BTRFS_INODE_COMPRESS;
323 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
325 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
329 ret = btrfs_set_prop(inode, "btrfs.compression",
330 comp, strlen(comp), 0);
335 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
338 trans = btrfs_start_transaction(root, 1);
340 ret = PTR_ERR(trans);
344 btrfs_update_iflags(inode);
345 inode_inc_iversion(inode);
346 inode->i_ctime = CURRENT_TIME;
347 ret = btrfs_update_inode(trans, root, inode);
349 btrfs_end_transaction(trans, root);
352 ip->flags = ip_oldflags;
353 inode->i_flags = i_oldflags;
357 mutex_unlock(&inode->i_mutex);
358 mnt_drop_write_file(file);
362 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
364 struct inode *inode = file_inode(file);
366 return put_user(inode->i_generation, arg);
369 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
371 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
372 struct btrfs_device *device;
373 struct request_queue *q;
374 struct fstrim_range range;
375 u64 minlen = ULLONG_MAX;
377 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
380 if (!capable(CAP_SYS_ADMIN))
384 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
388 q = bdev_get_queue(device->bdev);
389 if (blk_queue_discard(q)) {
391 minlen = min((u64)q->limits.discard_granularity,
399 if (copy_from_user(&range, arg, sizeof(range)))
401 if (range.start > total_bytes ||
402 range.len < fs_info->sb->s_blocksize)
405 range.len = min(range.len, total_bytes - range.start);
406 range.minlen = max(range.minlen, minlen);
407 ret = btrfs_trim_fs(fs_info->tree_root, &range);
411 if (copy_to_user(arg, &range, sizeof(range)))
417 int btrfs_is_empty_uuid(u8 *uuid)
421 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
428 static noinline int create_subvol(struct inode *dir,
429 struct dentry *dentry,
430 char *name, int namelen,
432 struct btrfs_qgroup_inherit *inherit)
434 struct btrfs_trans_handle *trans;
435 struct btrfs_key key;
436 struct btrfs_root_item root_item;
437 struct btrfs_inode_item *inode_item;
438 struct extent_buffer *leaf;
439 struct btrfs_root *root = BTRFS_I(dir)->root;
440 struct btrfs_root *new_root;
441 struct btrfs_block_rsv block_rsv;
442 struct timespec cur_time = CURRENT_TIME;
447 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
452 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
456 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
458 * The same as the snapshot creation, please see the comment
459 * of create_snapshot().
461 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
462 8, &qgroup_reserved, false);
466 trans = btrfs_start_transaction(root, 0);
468 ret = PTR_ERR(trans);
469 btrfs_subvolume_release_metadata(root, &block_rsv,
473 trans->block_rsv = &block_rsv;
474 trans->bytes_reserved = block_rsv.size;
476 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
480 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
481 0, objectid, NULL, 0, 0, 0);
487 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
488 btrfs_set_header_bytenr(leaf, leaf->start);
489 btrfs_set_header_generation(leaf, trans->transid);
490 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
491 btrfs_set_header_owner(leaf, objectid);
493 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
495 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
496 btrfs_header_chunk_tree_uuid(leaf),
498 btrfs_mark_buffer_dirty(leaf);
500 memset(&root_item, 0, sizeof(root_item));
502 inode_item = &root_item.inode;
503 btrfs_set_stack_inode_generation(inode_item, 1);
504 btrfs_set_stack_inode_size(inode_item, 3);
505 btrfs_set_stack_inode_nlink(inode_item, 1);
506 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
507 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
509 btrfs_set_root_flags(&root_item, 0);
510 btrfs_set_root_limit(&root_item, 0);
511 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
513 btrfs_set_root_bytenr(&root_item, leaf->start);
514 btrfs_set_root_generation(&root_item, trans->transid);
515 btrfs_set_root_level(&root_item, 0);
516 btrfs_set_root_refs(&root_item, 1);
517 btrfs_set_root_used(&root_item, leaf->len);
518 btrfs_set_root_last_snapshot(&root_item, 0);
520 btrfs_set_root_generation_v2(&root_item,
521 btrfs_root_generation(&root_item));
522 uuid_le_gen(&new_uuid);
523 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
524 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
525 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
526 root_item.ctime = root_item.otime;
527 btrfs_set_root_ctransid(&root_item, trans->transid);
528 btrfs_set_root_otransid(&root_item, trans->transid);
530 btrfs_tree_unlock(leaf);
531 free_extent_buffer(leaf);
534 btrfs_set_root_dirid(&root_item, new_dirid);
536 key.objectid = objectid;
538 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
539 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
544 key.offset = (u64)-1;
545 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
546 if (IS_ERR(new_root)) {
547 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
548 ret = PTR_ERR(new_root);
552 btrfs_record_root_in_trans(trans, new_root);
554 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
556 /* We potentially lose an unused inode item here */
557 btrfs_abort_transaction(trans, root, ret);
562 * insert the directory item
564 ret = btrfs_set_inode_index(dir, &index);
566 btrfs_abort_transaction(trans, root, ret);
570 ret = btrfs_insert_dir_item(trans, root,
571 name, namelen, dir, &key,
572 BTRFS_FT_DIR, index);
574 btrfs_abort_transaction(trans, root, ret);
578 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
579 ret = btrfs_update_inode(trans, root, dir);
582 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
583 objectid, root->root_key.objectid,
584 btrfs_ino(dir), index, name, namelen);
587 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
588 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
591 btrfs_abort_transaction(trans, root, ret);
594 trans->block_rsv = NULL;
595 trans->bytes_reserved = 0;
596 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
599 *async_transid = trans->transid;
600 err = btrfs_commit_transaction_async(trans, root, 1);
602 err = btrfs_commit_transaction(trans, root);
604 err = btrfs_commit_transaction(trans, root);
610 inode = btrfs_lookup_dentry(dir, dentry);
612 return PTR_ERR(inode);
613 d_instantiate(dentry, inode);
618 static void btrfs_wait_nocow_write(struct btrfs_root *root)
624 prepare_to_wait(&root->subv_writers->wait, &wait,
625 TASK_UNINTERRUPTIBLE);
627 writers = percpu_counter_sum(&root->subv_writers->counter);
631 finish_wait(&root->subv_writers->wait, &wait);
635 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
636 struct dentry *dentry, char *name, int namelen,
637 u64 *async_transid, bool readonly,
638 struct btrfs_qgroup_inherit *inherit)
641 struct btrfs_pending_snapshot *pending_snapshot;
642 struct btrfs_trans_handle *trans;
645 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
648 atomic_inc(&root->will_be_snapshoted);
649 smp_mb__after_atomic();
650 btrfs_wait_nocow_write(root);
652 ret = btrfs_start_delalloc_inodes(root, 0);
656 btrfs_wait_ordered_extents(root, -1);
658 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
659 if (!pending_snapshot) {
664 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
665 BTRFS_BLOCK_RSV_TEMP);
667 * 1 - parent dir inode
670 * 2 - root ref/backref
671 * 1 - root of snapshot
674 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
675 &pending_snapshot->block_rsv, 8,
676 &pending_snapshot->qgroup_reserved,
681 pending_snapshot->dentry = dentry;
682 pending_snapshot->root = root;
683 pending_snapshot->readonly = readonly;
684 pending_snapshot->dir = dir;
685 pending_snapshot->inherit = inherit;
687 trans = btrfs_start_transaction(root, 0);
689 ret = PTR_ERR(trans);
693 spin_lock(&root->fs_info->trans_lock);
694 list_add(&pending_snapshot->list,
695 &trans->transaction->pending_snapshots);
696 spin_unlock(&root->fs_info->trans_lock);
698 *async_transid = trans->transid;
699 ret = btrfs_commit_transaction_async(trans,
700 root->fs_info->extent_root, 1);
702 ret = btrfs_commit_transaction(trans, root);
704 ret = btrfs_commit_transaction(trans,
705 root->fs_info->extent_root);
710 ret = pending_snapshot->error;
714 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
716 ret = PTR_ERR(inode);
720 d_instantiate(dentry, inode);
723 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
724 &pending_snapshot->block_rsv,
725 pending_snapshot->qgroup_reserved);
727 kfree(pending_snapshot);
729 atomic_dec(&root->will_be_snapshoted);
733 /* copy of check_sticky in fs/namei.c()
734 * It's inline, so penalty for filesystems that don't use sticky bit is
737 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
739 kuid_t fsuid = current_fsuid();
741 if (!(dir->i_mode & S_ISVTX))
743 if (uid_eq(inode->i_uid, fsuid))
745 if (uid_eq(dir->i_uid, fsuid))
747 return !capable(CAP_FOWNER);
750 /* copy of may_delete in fs/namei.c()
751 * Check whether we can remove a link victim from directory dir, check
752 * whether the type of victim is right.
753 * 1. We can't do it if dir is read-only (done in permission())
754 * 2. We should have write and exec permissions on dir
755 * 3. We can't remove anything from append-only dir
756 * 4. We can't do anything with immutable dir (done in permission())
757 * 5. If the sticky bit on dir is set we should either
758 * a. be owner of dir, or
759 * b. be owner of victim, or
760 * c. have CAP_FOWNER capability
761 * 6. If the victim is append-only or immutable we can't do antyhing with
762 * links pointing to it.
763 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
764 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
765 * 9. We can't remove a root or mountpoint.
766 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
767 * nfs_async_unlink().
770 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
774 if (!victim->d_inode)
777 BUG_ON(victim->d_parent->d_inode != dir);
778 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
780 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
785 if (btrfs_check_sticky(dir, victim->d_inode)||
786 IS_APPEND(victim->d_inode)||
787 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
790 if (!S_ISDIR(victim->d_inode->i_mode))
794 } else if (S_ISDIR(victim->d_inode->i_mode))
798 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
803 /* copy of may_create in fs/namei.c() */
804 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
810 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
814 * Create a new subvolume below @parent. This is largely modeled after
815 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
816 * inside this filesystem so it's quite a bit simpler.
818 static noinline int btrfs_mksubvol(struct path *parent,
819 char *name, int namelen,
820 struct btrfs_root *snap_src,
821 u64 *async_transid, bool readonly,
822 struct btrfs_qgroup_inherit *inherit)
824 struct inode *dir = parent->dentry->d_inode;
825 struct dentry *dentry;
828 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
832 dentry = lookup_one_len(name, parent->dentry, namelen);
833 error = PTR_ERR(dentry);
841 error = btrfs_may_create(dir, dentry);
846 * even if this name doesn't exist, we may get hash collisions.
847 * check for them now when we can safely fail
849 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
855 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
857 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
861 error = create_snapshot(snap_src, dir, dentry, name, namelen,
862 async_transid, readonly, inherit);
864 error = create_subvol(dir, dentry, name, namelen,
865 async_transid, inherit);
868 fsnotify_mkdir(dir, dentry);
870 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
874 mutex_unlock(&dir->i_mutex);
879 * When we're defragging a range, we don't want to kick it off again
880 * if it is really just waiting for delalloc to send it down.
881 * If we find a nice big extent or delalloc range for the bytes in the
882 * file you want to defrag, we return 0 to let you know to skip this
885 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
887 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
888 struct extent_map *em = NULL;
889 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
892 read_lock(&em_tree->lock);
893 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
894 read_unlock(&em_tree->lock);
897 end = extent_map_end(em);
899 if (end - offset > thresh)
902 /* if we already have a nice delalloc here, just stop */
904 end = count_range_bits(io_tree, &offset, offset + thresh,
905 thresh, EXTENT_DELALLOC, 1);
912 * helper function to walk through a file and find extents
913 * newer than a specific transid, and smaller than thresh.
915 * This is used by the defragging code to find new and small
918 static int find_new_extents(struct btrfs_root *root,
919 struct inode *inode, u64 newer_than,
920 u64 *off, int thresh)
922 struct btrfs_path *path;
923 struct btrfs_key min_key;
924 struct extent_buffer *leaf;
925 struct btrfs_file_extent_item *extent;
928 u64 ino = btrfs_ino(inode);
930 path = btrfs_alloc_path();
934 min_key.objectid = ino;
935 min_key.type = BTRFS_EXTENT_DATA_KEY;
936 min_key.offset = *off;
939 path->keep_locks = 1;
940 ret = btrfs_search_forward(root, &min_key, path, newer_than);
943 path->keep_locks = 0;
944 btrfs_unlock_up_safe(path, 1);
946 if (min_key.objectid != ino)
948 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
951 leaf = path->nodes[0];
952 extent = btrfs_item_ptr(leaf, path->slots[0],
953 struct btrfs_file_extent_item);
955 type = btrfs_file_extent_type(leaf, extent);
956 if (type == BTRFS_FILE_EXTENT_REG &&
957 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
958 check_defrag_in_cache(inode, min_key.offset, thresh)) {
959 *off = min_key.offset;
960 btrfs_free_path(path);
965 if (path->slots[0] < btrfs_header_nritems(leaf)) {
966 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
970 if (min_key.offset == (u64)-1)
974 btrfs_release_path(path);
977 btrfs_free_path(path);
981 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
983 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
984 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
985 struct extent_map *em;
986 u64 len = PAGE_CACHE_SIZE;
989 * hopefully we have this extent in the tree already, try without
990 * the full extent lock
992 read_lock(&em_tree->lock);
993 em = lookup_extent_mapping(em_tree, start, len);
994 read_unlock(&em_tree->lock);
997 struct extent_state *cached = NULL;
998 u64 end = start + len - 1;
1000 /* get the big lock and read metadata off disk */
1001 lock_extent_bits(io_tree, start, end, 0, &cached);
1002 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
1003 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
1012 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1014 struct extent_map *next;
1017 /* this is the last extent */
1018 if (em->start + em->len >= i_size_read(inode))
1021 next = defrag_lookup_extent(inode, em->start + em->len);
1022 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1024 else if ((em->block_start + em->block_len == next->block_start) &&
1025 (em->block_len > 128 * 1024 && next->block_len > 128 * 1024))
1028 free_extent_map(next);
1032 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
1033 u64 *last_len, u64 *skip, u64 *defrag_end,
1036 struct extent_map *em;
1038 bool next_mergeable = true;
1041 * make sure that once we start defragging an extent, we keep on
1044 if (start < *defrag_end)
1049 em = defrag_lookup_extent(inode, start);
1053 /* this will cover holes, and inline extents */
1054 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1059 next_mergeable = defrag_check_next_extent(inode, em);
1061 * we hit a real extent, if it is big or the next extent is not a
1062 * real extent, don't bother defragging it
1064 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1065 (em->len >= thresh || !next_mergeable))
1069 * last_len ends up being a counter of how many bytes we've defragged.
1070 * every time we choose not to defrag an extent, we reset *last_len
1071 * so that the next tiny extent will force a defrag.
1073 * The end result of this is that tiny extents before a single big
1074 * extent will force at least part of that big extent to be defragged.
1077 *defrag_end = extent_map_end(em);
1080 *skip = extent_map_end(em);
1084 free_extent_map(em);
1089 * it doesn't do much good to defrag one or two pages
1090 * at a time. This pulls in a nice chunk of pages
1091 * to COW and defrag.
1093 * It also makes sure the delalloc code has enough
1094 * dirty data to avoid making new small extents as part
1097 * It's a good idea to start RA on this range
1098 * before calling this.
1100 static int cluster_pages_for_defrag(struct inode *inode,
1101 struct page **pages,
1102 unsigned long start_index,
1103 unsigned long num_pages)
1105 unsigned long file_end;
1106 u64 isize = i_size_read(inode);
1113 struct btrfs_ordered_extent *ordered;
1114 struct extent_state *cached_state = NULL;
1115 struct extent_io_tree *tree;
1116 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1118 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1119 if (!isize || start_index > file_end)
1122 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1124 ret = btrfs_delalloc_reserve_space(inode,
1125 page_cnt << PAGE_CACHE_SHIFT);
1129 tree = &BTRFS_I(inode)->io_tree;
1131 /* step one, lock all the pages */
1132 for (i = 0; i < page_cnt; i++) {
1135 page = find_or_create_page(inode->i_mapping,
1136 start_index + i, mask);
1140 page_start = page_offset(page);
1141 page_end = page_start + PAGE_CACHE_SIZE - 1;
1143 lock_extent_bits(tree, page_start, page_end,
1145 ordered = btrfs_lookup_ordered_extent(inode,
1147 unlock_extent_cached(tree, page_start, page_end,
1148 &cached_state, GFP_NOFS);
1153 btrfs_start_ordered_extent(inode, ordered, 1);
1154 btrfs_put_ordered_extent(ordered);
1157 * we unlocked the page above, so we need check if
1158 * it was released or not.
1160 if (page->mapping != inode->i_mapping) {
1162 page_cache_release(page);
1167 if (!PageUptodate(page)) {
1168 btrfs_readpage(NULL, page);
1170 if (!PageUptodate(page)) {
1172 page_cache_release(page);
1178 if (page->mapping != inode->i_mapping) {
1180 page_cache_release(page);
1190 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1194 * so now we have a nice long stream of locked
1195 * and up to date pages, lets wait on them
1197 for (i = 0; i < i_done; i++)
1198 wait_on_page_writeback(pages[i]);
1200 page_start = page_offset(pages[0]);
1201 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1203 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1204 page_start, page_end - 1, 0, &cached_state);
1205 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1206 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1207 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1208 &cached_state, GFP_NOFS);
1210 if (i_done != page_cnt) {
1211 spin_lock(&BTRFS_I(inode)->lock);
1212 BTRFS_I(inode)->outstanding_extents++;
1213 spin_unlock(&BTRFS_I(inode)->lock);
1214 btrfs_delalloc_release_space(inode,
1215 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1219 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1220 &cached_state, GFP_NOFS);
1222 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1223 page_start, page_end - 1, &cached_state,
1226 for (i = 0; i < i_done; i++) {
1227 clear_page_dirty_for_io(pages[i]);
1228 ClearPageChecked(pages[i]);
1229 set_page_extent_mapped(pages[i]);
1230 set_page_dirty(pages[i]);
1231 unlock_page(pages[i]);
1232 page_cache_release(pages[i]);
1236 for (i = 0; i < i_done; i++) {
1237 unlock_page(pages[i]);
1238 page_cache_release(pages[i]);
1240 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1245 int btrfs_defrag_file(struct inode *inode, struct file *file,
1246 struct btrfs_ioctl_defrag_range_args *range,
1247 u64 newer_than, unsigned long max_to_defrag)
1249 struct btrfs_root *root = BTRFS_I(inode)->root;
1250 struct file_ra_state *ra = NULL;
1251 unsigned long last_index;
1252 u64 isize = i_size_read(inode);
1256 u64 newer_off = range->start;
1258 unsigned long ra_index = 0;
1260 int defrag_count = 0;
1261 int compress_type = BTRFS_COMPRESS_ZLIB;
1262 int extent_thresh = range->extent_thresh;
1263 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1264 unsigned long cluster = max_cluster;
1265 u64 new_align = ~((u64)128 * 1024 - 1);
1266 struct page **pages = NULL;
1271 if (range->start >= isize)
1274 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1275 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1277 if (range->compress_type)
1278 compress_type = range->compress_type;
1281 if (extent_thresh == 0)
1282 extent_thresh = 256 * 1024;
1285 * if we were not given a file, allocate a readahead
1289 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1292 file_ra_state_init(ra, inode->i_mapping);
1297 pages = kmalloc_array(max_cluster, sizeof(struct page *),
1304 /* find the last page to defrag */
1305 if (range->start + range->len > range->start) {
1306 last_index = min_t(u64, isize - 1,
1307 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1309 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1313 ret = find_new_extents(root, inode, newer_than,
1314 &newer_off, 64 * 1024);
1316 range->start = newer_off;
1318 * we always align our defrag to help keep
1319 * the extents in the file evenly spaced
1321 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1325 i = range->start >> PAGE_CACHE_SHIFT;
1328 max_to_defrag = last_index + 1;
1331 * make writeback starts from i, so the defrag range can be
1332 * written sequentially.
1334 if (i < inode->i_mapping->writeback_index)
1335 inode->i_mapping->writeback_index = i;
1337 while (i <= last_index && defrag_count < max_to_defrag &&
1338 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1339 PAGE_CACHE_SHIFT)) {
1341 * make sure we stop running if someone unmounts
1344 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1347 if (btrfs_defrag_cancelled(root->fs_info)) {
1348 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
1353 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1354 extent_thresh, &last_len, &skip,
1355 &defrag_end, range->flags &
1356 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1359 * the should_defrag function tells us how much to skip
1360 * bump our counter by the suggested amount
1362 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1363 i = max(i + 1, next);
1368 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1369 PAGE_CACHE_SHIFT) - i;
1370 cluster = min(cluster, max_cluster);
1372 cluster = max_cluster;
1375 if (i + cluster > ra_index) {
1376 ra_index = max(i, ra_index);
1377 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1379 ra_index += max_cluster;
1382 mutex_lock(&inode->i_mutex);
1383 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1384 BTRFS_I(inode)->force_compress = compress_type;
1385 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1387 mutex_unlock(&inode->i_mutex);
1391 defrag_count += ret;
1392 balance_dirty_pages_ratelimited(inode->i_mapping);
1393 mutex_unlock(&inode->i_mutex);
1396 if (newer_off == (u64)-1)
1402 newer_off = max(newer_off + 1,
1403 (u64)i << PAGE_CACHE_SHIFT);
1405 ret = find_new_extents(root, inode,
1406 newer_than, &newer_off,
1409 range->start = newer_off;
1410 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1417 last_len += ret << PAGE_CACHE_SHIFT;
1425 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1426 filemap_flush(inode->i_mapping);
1427 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1428 &BTRFS_I(inode)->runtime_flags))
1429 filemap_flush(inode->i_mapping);
1432 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1433 /* the filemap_flush will queue IO into the worker threads, but
1434 * we have to make sure the IO is actually started and that
1435 * ordered extents get created before we return
1437 atomic_inc(&root->fs_info->async_submit_draining);
1438 while (atomic_read(&root->fs_info->nr_async_submits) ||
1439 atomic_read(&root->fs_info->async_delalloc_pages)) {
1440 wait_event(root->fs_info->async_submit_wait,
1441 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1442 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1444 atomic_dec(&root->fs_info->async_submit_draining);
1447 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1448 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1454 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1455 mutex_lock(&inode->i_mutex);
1456 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1457 mutex_unlock(&inode->i_mutex);
1465 static noinline int btrfs_ioctl_resize(struct file *file,
1471 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1472 struct btrfs_ioctl_vol_args *vol_args;
1473 struct btrfs_trans_handle *trans;
1474 struct btrfs_device *device = NULL;
1477 char *devstr = NULL;
1481 if (!capable(CAP_SYS_ADMIN))
1484 ret = mnt_want_write_file(file);
1488 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1490 mnt_drop_write_file(file);
1491 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1494 mutex_lock(&root->fs_info->volume_mutex);
1495 vol_args = memdup_user(arg, sizeof(*vol_args));
1496 if (IS_ERR(vol_args)) {
1497 ret = PTR_ERR(vol_args);
1501 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1503 sizestr = vol_args->name;
1504 devstr = strchr(sizestr, ':');
1506 sizestr = devstr + 1;
1508 devstr = vol_args->name;
1509 ret = kstrtoull(devstr, 10, &devid);
1516 btrfs_info(root->fs_info, "resizing devid %llu", devid);
1519 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1521 btrfs_info(root->fs_info, "resizer unable to find device %llu",
1527 if (!device->writeable) {
1528 btrfs_info(root->fs_info,
1529 "resizer unable to apply on readonly device %llu",
1535 if (!strcmp(sizestr, "max"))
1536 new_size = device->bdev->bd_inode->i_size;
1538 if (sizestr[0] == '-') {
1541 } else if (sizestr[0] == '+') {
1545 new_size = memparse(sizestr, &retptr);
1546 if (*retptr != '\0' || new_size == 0) {
1552 if (device->is_tgtdev_for_dev_replace) {
1557 old_size = device->total_bytes;
1560 if (new_size > old_size) {
1564 new_size = old_size - new_size;
1565 } else if (mod > 0) {
1566 if (new_size > ULLONG_MAX - old_size) {
1570 new_size = old_size + new_size;
1573 if (new_size < 256 * 1024 * 1024) {
1577 if (new_size > device->bdev->bd_inode->i_size) {
1582 do_div(new_size, root->sectorsize);
1583 new_size *= root->sectorsize;
1585 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
1586 rcu_str_deref(device->name), new_size);
1588 if (new_size > old_size) {
1589 trans = btrfs_start_transaction(root, 0);
1590 if (IS_ERR(trans)) {
1591 ret = PTR_ERR(trans);
1594 ret = btrfs_grow_device(trans, device, new_size);
1595 btrfs_commit_transaction(trans, root);
1596 } else if (new_size < old_size) {
1597 ret = btrfs_shrink_device(device, new_size);
1598 } /* equal, nothing need to do */
1603 mutex_unlock(&root->fs_info->volume_mutex);
1604 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1605 mnt_drop_write_file(file);
1609 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1610 char *name, unsigned long fd, int subvol,
1611 u64 *transid, bool readonly,
1612 struct btrfs_qgroup_inherit *inherit)
1617 ret = mnt_want_write_file(file);
1621 namelen = strlen(name);
1622 if (strchr(name, '/')) {
1624 goto out_drop_write;
1627 if (name[0] == '.' &&
1628 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1630 goto out_drop_write;
1634 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1635 NULL, transid, readonly, inherit);
1637 struct fd src = fdget(fd);
1638 struct inode *src_inode;
1641 goto out_drop_write;
1644 src_inode = file_inode(src.file);
1645 if (src_inode->i_sb != file_inode(file)->i_sb) {
1646 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1647 "Snapshot src from another FS");
1649 } else if (!inode_owner_or_capable(src_inode)) {
1651 * Subvolume creation is not restricted, but snapshots
1652 * are limited to own subvolumes only
1656 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1657 BTRFS_I(src_inode)->root,
1658 transid, readonly, inherit);
1663 mnt_drop_write_file(file);
1668 static noinline int btrfs_ioctl_snap_create(struct file *file,
1669 void __user *arg, int subvol)
1671 struct btrfs_ioctl_vol_args *vol_args;
1674 vol_args = memdup_user(arg, sizeof(*vol_args));
1675 if (IS_ERR(vol_args))
1676 return PTR_ERR(vol_args);
1677 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1679 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1680 vol_args->fd, subvol,
1687 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1688 void __user *arg, int subvol)
1690 struct btrfs_ioctl_vol_args_v2 *vol_args;
1694 bool readonly = false;
1695 struct btrfs_qgroup_inherit *inherit = NULL;
1697 vol_args = memdup_user(arg, sizeof(*vol_args));
1698 if (IS_ERR(vol_args))
1699 return PTR_ERR(vol_args);
1700 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1702 if (vol_args->flags &
1703 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1704 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1709 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1711 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1713 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1714 if (vol_args->size > PAGE_CACHE_SIZE) {
1718 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1719 if (IS_ERR(inherit)) {
1720 ret = PTR_ERR(inherit);
1725 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1726 vol_args->fd, subvol, ptr,
1731 if (ptr && copy_to_user(arg +
1732 offsetof(struct btrfs_ioctl_vol_args_v2,
1744 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1747 struct inode *inode = file_inode(file);
1748 struct btrfs_root *root = BTRFS_I(inode)->root;
1752 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1755 down_read(&root->fs_info->subvol_sem);
1756 if (btrfs_root_readonly(root))
1757 flags |= BTRFS_SUBVOL_RDONLY;
1758 up_read(&root->fs_info->subvol_sem);
1760 if (copy_to_user(arg, &flags, sizeof(flags)))
1766 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1769 struct inode *inode = file_inode(file);
1770 struct btrfs_root *root = BTRFS_I(inode)->root;
1771 struct btrfs_trans_handle *trans;
1776 if (!inode_owner_or_capable(inode))
1779 ret = mnt_want_write_file(file);
1783 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1785 goto out_drop_write;
1788 if (copy_from_user(&flags, arg, sizeof(flags))) {
1790 goto out_drop_write;
1793 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1795 goto out_drop_write;
1798 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1800 goto out_drop_write;
1803 down_write(&root->fs_info->subvol_sem);
1806 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1809 root_flags = btrfs_root_flags(&root->root_item);
1810 if (flags & BTRFS_SUBVOL_RDONLY) {
1811 btrfs_set_root_flags(&root->root_item,
1812 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1815 * Block RO -> RW transition if this subvolume is involved in
1818 spin_lock(&root->root_item_lock);
1819 if (root->send_in_progress == 0) {
1820 btrfs_set_root_flags(&root->root_item,
1821 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1822 spin_unlock(&root->root_item_lock);
1824 spin_unlock(&root->root_item_lock);
1825 btrfs_warn(root->fs_info,
1826 "Attempt to set subvolume %llu read-write during send",
1827 root->root_key.objectid);
1833 trans = btrfs_start_transaction(root, 1);
1834 if (IS_ERR(trans)) {
1835 ret = PTR_ERR(trans);
1839 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1840 &root->root_key, &root->root_item);
1842 btrfs_commit_transaction(trans, root);
1845 btrfs_set_root_flags(&root->root_item, root_flags);
1847 up_write(&root->fs_info->subvol_sem);
1849 mnt_drop_write_file(file);
1855 * helper to check if the subvolume references other subvolumes
1857 static noinline int may_destroy_subvol(struct btrfs_root *root)
1859 struct btrfs_path *path;
1860 struct btrfs_dir_item *di;
1861 struct btrfs_key key;
1865 path = btrfs_alloc_path();
1869 /* Make sure this root isn't set as the default subvol */
1870 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1871 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1872 dir_id, "default", 7, 0);
1873 if (di && !IS_ERR(di)) {
1874 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1875 if (key.objectid == root->root_key.objectid) {
1877 btrfs_err(root->fs_info, "deleting default subvolume "
1878 "%llu is not allowed", key.objectid);
1881 btrfs_release_path(path);
1884 key.objectid = root->root_key.objectid;
1885 key.type = BTRFS_ROOT_REF_KEY;
1886 key.offset = (u64)-1;
1888 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1895 if (path->slots[0] > 0) {
1897 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1898 if (key.objectid == root->root_key.objectid &&
1899 key.type == BTRFS_ROOT_REF_KEY)
1903 btrfs_free_path(path);
1907 static noinline int key_in_sk(struct btrfs_key *key,
1908 struct btrfs_ioctl_search_key *sk)
1910 struct btrfs_key test;
1913 test.objectid = sk->min_objectid;
1914 test.type = sk->min_type;
1915 test.offset = sk->min_offset;
1917 ret = btrfs_comp_cpu_keys(key, &test);
1921 test.objectid = sk->max_objectid;
1922 test.type = sk->max_type;
1923 test.offset = sk->max_offset;
1925 ret = btrfs_comp_cpu_keys(key, &test);
1931 static noinline int copy_to_sk(struct btrfs_root *root,
1932 struct btrfs_path *path,
1933 struct btrfs_key *key,
1934 struct btrfs_ioctl_search_key *sk,
1937 unsigned long *sk_offset,
1941 struct extent_buffer *leaf;
1942 struct btrfs_ioctl_search_header sh;
1943 unsigned long item_off;
1944 unsigned long item_len;
1950 leaf = path->nodes[0];
1951 slot = path->slots[0];
1952 nritems = btrfs_header_nritems(leaf);
1954 if (btrfs_header_generation(leaf) > sk->max_transid) {
1958 found_transid = btrfs_header_generation(leaf);
1960 for (i = slot; i < nritems; i++) {
1961 item_off = btrfs_item_ptr_offset(leaf, i);
1962 item_len = btrfs_item_size_nr(leaf, i);
1964 btrfs_item_key_to_cpu(leaf, key, i);
1965 if (!key_in_sk(key, sk))
1968 if (sizeof(sh) + item_len > *buf_size) {
1975 * return one empty item back for v1, which does not
1979 *buf_size = sizeof(sh) + item_len;
1984 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1989 sh.objectid = key->objectid;
1990 sh.offset = key->offset;
1991 sh.type = key->type;
1993 sh.transid = found_transid;
1995 /* copy search result header */
1996 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2001 *sk_offset += sizeof(sh);
2004 char __user *up = ubuf + *sk_offset;
2006 if (read_extent_buffer_to_user(leaf, up,
2007 item_off, item_len)) {
2012 *sk_offset += item_len;
2016 if (ret) /* -EOVERFLOW from above */
2019 if (*num_found >= sk->nr_items) {
2026 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
2028 else if (key->type < (u8)-1 && key->type < sk->max_type) {
2031 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
2039 * 0: all items from this leaf copied, continue with next
2040 * 1: * more items can be copied, but unused buffer is too small
2041 * * all items were found
2042 * Either way, it will stops the loop which iterates to the next
2044 * -EOVERFLOW: item was to large for buffer
2045 * -EFAULT: could not copy extent buffer back to userspace
2050 static noinline int search_ioctl(struct inode *inode,
2051 struct btrfs_ioctl_search_key *sk,
2055 struct btrfs_root *root;
2056 struct btrfs_key key;
2057 struct btrfs_path *path;
2058 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2061 unsigned long sk_offset = 0;
2063 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2064 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2068 path = btrfs_alloc_path();
2072 if (sk->tree_id == 0) {
2073 /* search the root of the inode that was passed */
2074 root = BTRFS_I(inode)->root;
2076 key.objectid = sk->tree_id;
2077 key.type = BTRFS_ROOT_ITEM_KEY;
2078 key.offset = (u64)-1;
2079 root = btrfs_read_fs_root_no_name(info, &key);
2081 printk(KERN_ERR "BTRFS: could not find root %llu\n",
2083 btrfs_free_path(path);
2088 key.objectid = sk->min_objectid;
2089 key.type = sk->min_type;
2090 key.offset = sk->min_offset;
2092 path->keep_locks = 1;
2095 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2101 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
2102 &sk_offset, &num_found);
2103 btrfs_release_path(path);
2111 sk->nr_items = num_found;
2112 btrfs_free_path(path);
2116 static noinline int btrfs_ioctl_tree_search(struct file *file,
2119 struct btrfs_ioctl_search_args __user *uargs;
2120 struct btrfs_ioctl_search_key sk;
2121 struct inode *inode;
2125 if (!capable(CAP_SYS_ADMIN))
2128 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2130 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2133 buf_size = sizeof(uargs->buf);
2135 inode = file_inode(file);
2136 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2139 * In the origin implementation an overflow is handled by returning a
2140 * search header with a len of zero, so reset ret.
2142 if (ret == -EOVERFLOW)
2145 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2150 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2153 struct btrfs_ioctl_search_args_v2 __user *uarg;
2154 struct btrfs_ioctl_search_args_v2 args;
2155 struct inode *inode;
2158 const size_t buf_limit = 16 * 1024 * 1024;
2160 if (!capable(CAP_SYS_ADMIN))
2163 /* copy search header and buffer size */
2164 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2165 if (copy_from_user(&args, uarg, sizeof(args)))
2168 buf_size = args.buf_size;
2170 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2173 /* limit result size to 16MB */
2174 if (buf_size > buf_limit)
2175 buf_size = buf_limit;
2177 inode = file_inode(file);
2178 ret = search_ioctl(inode, &args.key, &buf_size,
2179 (char *)(&uarg->buf[0]));
2180 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2182 else if (ret == -EOVERFLOW &&
2183 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2190 * Search INODE_REFs to identify path name of 'dirid' directory
2191 * in a 'tree_id' tree. and sets path name to 'name'.
2193 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2194 u64 tree_id, u64 dirid, char *name)
2196 struct btrfs_root *root;
2197 struct btrfs_key key;
2203 struct btrfs_inode_ref *iref;
2204 struct extent_buffer *l;
2205 struct btrfs_path *path;
2207 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2212 path = btrfs_alloc_path();
2216 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2218 key.objectid = tree_id;
2219 key.type = BTRFS_ROOT_ITEM_KEY;
2220 key.offset = (u64)-1;
2221 root = btrfs_read_fs_root_no_name(info, &key);
2223 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
2228 key.objectid = dirid;
2229 key.type = BTRFS_INODE_REF_KEY;
2230 key.offset = (u64)-1;
2233 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2237 ret = btrfs_previous_item(root, path, dirid,
2238 BTRFS_INODE_REF_KEY);
2248 slot = path->slots[0];
2249 btrfs_item_key_to_cpu(l, &key, slot);
2251 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2252 len = btrfs_inode_ref_name_len(l, iref);
2254 total_len += len + 1;
2256 ret = -ENAMETOOLONG;
2261 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2263 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2266 btrfs_release_path(path);
2267 key.objectid = key.offset;
2268 key.offset = (u64)-1;
2269 dirid = key.objectid;
2271 memmove(name, ptr, total_len);
2272 name[total_len] = '\0';
2275 btrfs_free_path(path);
2279 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2282 struct btrfs_ioctl_ino_lookup_args *args;
2283 struct inode *inode;
2286 if (!capable(CAP_SYS_ADMIN))
2289 args = memdup_user(argp, sizeof(*args));
2291 return PTR_ERR(args);
2293 inode = file_inode(file);
2295 if (args->treeid == 0)
2296 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2298 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2299 args->treeid, args->objectid,
2302 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2309 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2312 struct dentry *parent = file->f_path.dentry;
2313 struct dentry *dentry;
2314 struct inode *dir = parent->d_inode;
2315 struct inode *inode;
2316 struct btrfs_root *root = BTRFS_I(dir)->root;
2317 struct btrfs_root *dest = NULL;
2318 struct btrfs_ioctl_vol_args *vol_args;
2319 struct btrfs_trans_handle *trans;
2320 struct btrfs_block_rsv block_rsv;
2322 u64 qgroup_reserved;
2327 vol_args = memdup_user(arg, sizeof(*vol_args));
2328 if (IS_ERR(vol_args))
2329 return PTR_ERR(vol_args);
2331 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2332 namelen = strlen(vol_args->name);
2333 if (strchr(vol_args->name, '/') ||
2334 strncmp(vol_args->name, "..", namelen) == 0) {
2339 err = mnt_want_write_file(file);
2344 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2346 goto out_drop_write;
2347 dentry = lookup_one_len(vol_args->name, parent, namelen);
2348 if (IS_ERR(dentry)) {
2349 err = PTR_ERR(dentry);
2350 goto out_unlock_dir;
2353 if (!dentry->d_inode) {
2358 inode = dentry->d_inode;
2359 dest = BTRFS_I(inode)->root;
2360 if (!capable(CAP_SYS_ADMIN)) {
2362 * Regular user. Only allow this with a special mount
2363 * option, when the user has write+exec access to the
2364 * subvol root, and when rmdir(2) would have been
2367 * Note that this is _not_ check that the subvol is
2368 * empty or doesn't contain data that we wouldn't
2369 * otherwise be able to delete.
2371 * Users who want to delete empty subvols should try
2375 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2379 * Do not allow deletion if the parent dir is the same
2380 * as the dir to be deleted. That means the ioctl
2381 * must be called on the dentry referencing the root
2382 * of the subvol, not a random directory contained
2389 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2394 /* check if subvolume may be deleted by a user */
2395 err = btrfs_may_delete(dir, dentry, 1);
2399 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2404 mutex_lock(&inode->i_mutex);
2407 * Don't allow to delete a subvolume with send in progress. This is
2408 * inside the i_mutex so the error handling that has to drop the bit
2409 * again is not run concurrently.
2411 spin_lock(&dest->root_item_lock);
2412 root_flags = btrfs_root_flags(&dest->root_item);
2413 if (dest->send_in_progress == 0) {
2414 btrfs_set_root_flags(&dest->root_item,
2415 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2416 spin_unlock(&dest->root_item_lock);
2418 spin_unlock(&dest->root_item_lock);
2419 btrfs_warn(root->fs_info,
2420 "Attempt to delete subvolume %llu during send",
2421 dest->root_key.objectid);
2426 err = d_invalidate(dentry);
2430 down_write(&root->fs_info->subvol_sem);
2432 err = may_destroy_subvol(dest);
2436 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2438 * One for dir inode, two for dir entries, two for root
2441 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2442 5, &qgroup_reserved, true);
2446 trans = btrfs_start_transaction(root, 0);
2447 if (IS_ERR(trans)) {
2448 err = PTR_ERR(trans);
2451 trans->block_rsv = &block_rsv;
2452 trans->bytes_reserved = block_rsv.size;
2454 ret = btrfs_unlink_subvol(trans, root, dir,
2455 dest->root_key.objectid,
2456 dentry->d_name.name,
2457 dentry->d_name.len);
2460 btrfs_abort_transaction(trans, root, ret);
2464 btrfs_record_root_in_trans(trans, dest);
2466 memset(&dest->root_item.drop_progress, 0,
2467 sizeof(dest->root_item.drop_progress));
2468 dest->root_item.drop_level = 0;
2469 btrfs_set_root_refs(&dest->root_item, 0);
2471 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2472 ret = btrfs_insert_orphan_item(trans,
2473 root->fs_info->tree_root,
2474 dest->root_key.objectid);
2476 btrfs_abort_transaction(trans, root, ret);
2482 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2483 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2484 dest->root_key.objectid);
2485 if (ret && ret != -ENOENT) {
2486 btrfs_abort_transaction(trans, root, ret);
2490 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2491 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2492 dest->root_item.received_uuid,
2493 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2494 dest->root_key.objectid);
2495 if (ret && ret != -ENOENT) {
2496 btrfs_abort_transaction(trans, root, ret);
2503 trans->block_rsv = NULL;
2504 trans->bytes_reserved = 0;
2505 ret = btrfs_end_transaction(trans, root);
2508 inode->i_flags |= S_DEAD;
2510 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2512 up_write(&root->fs_info->subvol_sem);
2515 spin_lock(&dest->root_item_lock);
2516 root_flags = btrfs_root_flags(&dest->root_item);
2517 btrfs_set_root_flags(&dest->root_item,
2518 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2519 spin_unlock(&dest->root_item_lock);
2521 mutex_unlock(&inode->i_mutex);
2523 shrink_dcache_sb(root->fs_info->sb);
2524 btrfs_invalidate_inodes(dest);
2526 ASSERT(dest->send_in_progress == 0);
2529 if (dest->cache_inode) {
2530 iput(dest->cache_inode);
2531 dest->cache_inode = NULL;
2537 mutex_unlock(&dir->i_mutex);
2539 mnt_drop_write_file(file);
2545 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2547 struct inode *inode = file_inode(file);
2548 struct btrfs_root *root = BTRFS_I(inode)->root;
2549 struct btrfs_ioctl_defrag_range_args *range;
2552 ret = mnt_want_write_file(file);
2556 if (btrfs_root_readonly(root)) {
2561 switch (inode->i_mode & S_IFMT) {
2563 if (!capable(CAP_SYS_ADMIN)) {
2567 ret = btrfs_defrag_root(root);
2570 ret = btrfs_defrag_root(root->fs_info->extent_root);
2573 if (!(file->f_mode & FMODE_WRITE)) {
2578 range = kzalloc(sizeof(*range), GFP_KERNEL);
2585 if (copy_from_user(range, argp,
2591 /* compression requires us to start the IO */
2592 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2593 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2594 range->extent_thresh = (u32)-1;
2597 /* the rest are all set to zero by kzalloc */
2598 range->len = (u64)-1;
2600 ret = btrfs_defrag_file(file_inode(file), file,
2610 mnt_drop_write_file(file);
2614 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2616 struct btrfs_ioctl_vol_args *vol_args;
2619 if (!capable(CAP_SYS_ADMIN))
2622 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2624 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2627 mutex_lock(&root->fs_info->volume_mutex);
2628 vol_args = memdup_user(arg, sizeof(*vol_args));
2629 if (IS_ERR(vol_args)) {
2630 ret = PTR_ERR(vol_args);
2634 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2635 ret = btrfs_init_new_device(root, vol_args->name);
2639 mutex_unlock(&root->fs_info->volume_mutex);
2640 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2644 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2646 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2647 struct btrfs_ioctl_vol_args *vol_args;
2650 if (!capable(CAP_SYS_ADMIN))
2653 ret = mnt_want_write_file(file);
2657 vol_args = memdup_user(arg, sizeof(*vol_args));
2658 if (IS_ERR(vol_args)) {
2659 ret = PTR_ERR(vol_args);
2663 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2665 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2667 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2671 mutex_lock(&root->fs_info->volume_mutex);
2672 ret = btrfs_rm_device(root, vol_args->name);
2673 mutex_unlock(&root->fs_info->volume_mutex);
2674 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2679 mnt_drop_write_file(file);
2683 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2685 struct btrfs_ioctl_fs_info_args *fi_args;
2686 struct btrfs_device *device;
2687 struct btrfs_device *next;
2688 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2691 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2695 mutex_lock(&fs_devices->device_list_mutex);
2696 fi_args->num_devices = fs_devices->num_devices;
2697 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2699 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2700 if (device->devid > fi_args->max_id)
2701 fi_args->max_id = device->devid;
2703 mutex_unlock(&fs_devices->device_list_mutex);
2705 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2706 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2707 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2709 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2716 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2718 struct btrfs_ioctl_dev_info_args *di_args;
2719 struct btrfs_device *dev;
2720 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2722 char *s_uuid = NULL;
2724 di_args = memdup_user(arg, sizeof(*di_args));
2725 if (IS_ERR(di_args))
2726 return PTR_ERR(di_args);
2728 if (!btrfs_is_empty_uuid(di_args->uuid))
2729 s_uuid = di_args->uuid;
2731 mutex_lock(&fs_devices->device_list_mutex);
2732 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2739 di_args->devid = dev->devid;
2740 di_args->bytes_used = dev->bytes_used;
2741 di_args->total_bytes = dev->total_bytes;
2742 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2744 struct rcu_string *name;
2747 name = rcu_dereference(dev->name);
2748 strncpy(di_args->path, name->str, sizeof(di_args->path));
2750 di_args->path[sizeof(di_args->path) - 1] = 0;
2752 di_args->path[0] = '\0';
2756 mutex_unlock(&fs_devices->device_list_mutex);
2757 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2764 static struct page *extent_same_get_page(struct inode *inode, u64 off)
2768 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2770 index = off >> PAGE_CACHE_SHIFT;
2772 page = grab_cache_page(inode->i_mapping, index);
2776 if (!PageUptodate(page)) {
2777 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2781 if (!PageUptodate(page)) {
2783 page_cache_release(page);
2792 static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2794 /* do any pending delalloc/csum calc on src, one way or
2795 another, and lock file content */
2797 struct btrfs_ordered_extent *ordered;
2798 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2799 ordered = btrfs_lookup_first_ordered_extent(inode,
2802 ordered->file_offset + ordered->len <= off ||
2803 ordered->file_offset >= off + len) &&
2804 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2805 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2807 btrfs_put_ordered_extent(ordered);
2810 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2812 btrfs_put_ordered_extent(ordered);
2813 btrfs_wait_ordered_range(inode, off, len);
2817 static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2818 struct inode *inode2, u64 loff2, u64 len)
2820 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2821 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2823 mutex_unlock(&inode1->i_mutex);
2824 mutex_unlock(&inode2->i_mutex);
2827 static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2828 struct inode *inode2, u64 loff2, u64 len)
2830 if (inode1 < inode2) {
2831 swap(inode1, inode2);
2835 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2836 lock_extent_range(inode1, loff1, len);
2837 if (inode1 != inode2) {
2838 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2839 lock_extent_range(inode2, loff2, len);
2843 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2844 u64 dst_loff, u64 len)
2847 struct page *src_page, *dst_page;
2848 unsigned int cmp_len = PAGE_CACHE_SIZE;
2849 void *addr, *dst_addr;
2852 if (len < PAGE_CACHE_SIZE)
2855 src_page = extent_same_get_page(src, loff);
2858 dst_page = extent_same_get_page(dst, dst_loff);
2860 page_cache_release(src_page);
2863 addr = kmap_atomic(src_page);
2864 dst_addr = kmap_atomic(dst_page);
2866 flush_dcache_page(src_page);
2867 flush_dcache_page(dst_page);
2869 if (memcmp(addr, dst_addr, cmp_len))
2870 ret = BTRFS_SAME_DATA_DIFFERS;
2872 kunmap_atomic(addr);
2873 kunmap_atomic(dst_addr);
2874 page_cache_release(src_page);
2875 page_cache_release(dst_page);
2881 dst_loff += cmp_len;
2888 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2890 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2892 if (off + len > inode->i_size || off + len < off)
2894 /* Check that we are block aligned - btrfs_clone() requires this */
2895 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2901 static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2902 struct inode *dst, u64 dst_loff)
2907 * btrfs_clone() can't handle extents in the same file
2908 * yet. Once that works, we can drop this check and replace it
2909 * with a check for the same inode, but overlapping extents.
2914 btrfs_double_lock(src, loff, dst, dst_loff, len);
2916 ret = extent_same_check_offsets(src, loff, len);
2920 ret = extent_same_check_offsets(dst, dst_loff, len);
2924 /* don't make the dst file partly checksummed */
2925 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2926 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2931 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2933 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2936 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2941 #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2943 static long btrfs_ioctl_file_extent_same(struct file *file,
2944 struct btrfs_ioctl_same_args __user *argp)
2946 struct btrfs_ioctl_same_args *same;
2947 struct btrfs_ioctl_same_extent_info *info;
2948 struct inode *src = file_inode(file);
2954 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2955 bool is_admin = capable(CAP_SYS_ADMIN);
2958 if (!(file->f_mode & FMODE_READ))
2961 ret = mnt_want_write_file(file);
2965 if (get_user(count, &argp->dest_count)) {
2970 size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
2972 same = memdup_user(argp, size);
2975 ret = PTR_ERR(same);
2979 off = same->logical_offset;
2983 * Limit the total length we will dedupe for each operation.
2984 * This is intended to bound the total time spent in this
2985 * ioctl to something sane.
2987 if (len > BTRFS_MAX_DEDUPE_LEN)
2988 len = BTRFS_MAX_DEDUPE_LEN;
2990 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
2992 * Btrfs does not support blocksize < page_size. As a
2993 * result, btrfs_cmp_data() won't correctly handle
2994 * this situation without an update.
3001 if (S_ISDIR(src->i_mode))
3005 if (!S_ISREG(src->i_mode))
3008 /* pre-format output fields to sane values */
3009 for (i = 0; i < count; i++) {
3010 same->info[i].bytes_deduped = 0ULL;
3011 same->info[i].status = 0;
3014 for (i = 0, info = same->info; i < count; i++, info++) {
3016 struct fd dst_file = fdget(info->fd);
3017 if (!dst_file.file) {
3018 info->status = -EBADF;
3021 dst = file_inode(dst_file.file);
3023 if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
3024 info->status = -EINVAL;
3025 } else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
3026 info->status = -EXDEV;
3027 } else if (S_ISDIR(dst->i_mode)) {
3028 info->status = -EISDIR;
3029 } else if (!S_ISREG(dst->i_mode)) {
3030 info->status = -EACCES;
3032 info->status = btrfs_extent_same(src, off, len, dst,
3033 info->logical_offset);
3034 if (info->status == 0)
3035 info->bytes_deduped += len;
3040 ret = copy_to_user(argp, same, size);
3045 mnt_drop_write_file(file);
3049 /* Helper to check and see if this root currently has a ref on the given disk
3050 * bytenr. If it does then we need to update the quota for this root. This
3051 * doesn't do anything if quotas aren't enabled.
3053 static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
3056 struct seq_list tree_mod_seq_elem = {};
3057 struct ulist *roots;
3058 struct ulist_iterator uiter;
3059 struct ulist_node *root_node = NULL;
3062 if (!root->fs_info->quota_enabled)
3065 btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3066 ret = btrfs_find_all_roots(trans, root->fs_info, disko,
3067 tree_mod_seq_elem.seq, &roots);
3071 ULIST_ITER_INIT(&uiter);
3072 while ((root_node = ulist_next(roots, &uiter))) {
3073 if (root_node->val == root->objectid) {
3080 btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3084 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3085 struct inode *inode,
3090 struct btrfs_root *root = BTRFS_I(inode)->root;
3093 inode_inc_iversion(inode);
3094 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3096 * We round up to the block size at eof when determining which
3097 * extents to clone above, but shouldn't round up the file size.
3099 if (endoff > destoff + olen)
3100 endoff = destoff + olen;
3101 if (endoff > inode->i_size)
3102 btrfs_i_size_write(inode, endoff);
3104 ret = btrfs_update_inode(trans, root, inode);
3106 btrfs_abort_transaction(trans, root, ret);
3107 btrfs_end_transaction(trans, root);
3110 ret = btrfs_end_transaction(trans, root);
3115 static void clone_update_extent_map(struct inode *inode,
3116 const struct btrfs_trans_handle *trans,
3117 const struct btrfs_path *path,
3118 const u64 hole_offset,
3121 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3122 struct extent_map *em;
3125 em = alloc_extent_map();
3127 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3128 &BTRFS_I(inode)->runtime_flags);
3133 struct btrfs_file_extent_item *fi;
3135 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3136 struct btrfs_file_extent_item);
3137 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3138 em->generation = -1;
3139 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3140 BTRFS_FILE_EXTENT_INLINE)
3141 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3142 &BTRFS_I(inode)->runtime_flags);
3144 em->start = hole_offset;
3146 em->ram_bytes = em->len;
3147 em->orig_start = hole_offset;
3148 em->block_start = EXTENT_MAP_HOLE;
3150 em->orig_block_len = 0;
3151 em->compress_type = BTRFS_COMPRESS_NONE;
3152 em->generation = trans->transid;
3156 write_lock(&em_tree->lock);
3157 ret = add_extent_mapping(em_tree, em, 1);
3158 write_unlock(&em_tree->lock);
3159 if (ret != -EEXIST) {
3160 free_extent_map(em);
3163 btrfs_drop_extent_cache(inode, em->start,
3164 em->start + em->len - 1, 0);
3168 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3169 &BTRFS_I(inode)->runtime_flags);
3173 * btrfs_clone() - clone a range from inode file to another
3175 * @src: Inode to clone from
3176 * @inode: Inode to clone to
3177 * @off: Offset within source to start clone from
3178 * @olen: Original length, passed by user, of range to clone
3179 * @olen_aligned: Block-aligned value of olen, extent_same uses
3180 * identical values here
3181 * @destoff: Offset within @inode to start clone
3183 static int btrfs_clone(struct inode *src, struct inode *inode,
3184 const u64 off, const u64 olen, const u64 olen_aligned,
3187 struct btrfs_root *root = BTRFS_I(inode)->root;
3188 struct btrfs_path *path = NULL;
3189 struct extent_buffer *leaf;
3190 struct btrfs_trans_handle *trans;
3192 struct btrfs_key key;
3197 const u64 len = olen_aligned;
3199 u64 last_dest_end = destoff;
3202 buf = vmalloc(btrfs_level_size(root, 0));
3206 path = btrfs_alloc_path();
3214 key.objectid = btrfs_ino(src);
3215 key.type = BTRFS_EXTENT_DATA_KEY;
3220 * note the key will change type as we walk through the
3223 path->leave_spinning = 1;
3224 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3229 * First search, if no extent item that starts at offset off was
3230 * found but the previous item is an extent item, it's possible
3231 * it might overlap our target range, therefore process it.
3233 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3234 btrfs_item_key_to_cpu(path->nodes[0], &key,
3235 path->slots[0] - 1);
3236 if (key.type == BTRFS_EXTENT_DATA_KEY)
3240 nritems = btrfs_header_nritems(path->nodes[0]);
3243 if (path->slots[0] >= nritems) {
3244 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3249 nritems = btrfs_header_nritems(path->nodes[0]);
3251 leaf = path->nodes[0];
3252 slot = path->slots[0];
3254 btrfs_item_key_to_cpu(leaf, &key, slot);
3255 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
3256 key.objectid != btrfs_ino(src))
3259 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
3260 struct btrfs_file_extent_item *extent;
3263 struct btrfs_key new_key;
3264 u64 disko = 0, diskl = 0;
3265 u64 datao = 0, datal = 0;
3269 extent = btrfs_item_ptr(leaf, slot,
3270 struct btrfs_file_extent_item);
3271 comp = btrfs_file_extent_compression(leaf, extent);
3272 type = btrfs_file_extent_type(leaf, extent);
3273 if (type == BTRFS_FILE_EXTENT_REG ||
3274 type == BTRFS_FILE_EXTENT_PREALLOC) {
3275 disko = btrfs_file_extent_disk_bytenr(leaf,
3277 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3279 datao = btrfs_file_extent_offset(leaf, extent);
3280 datal = btrfs_file_extent_num_bytes(leaf,
3282 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3283 /* take upper bound, may be compressed */
3284 datal = btrfs_file_extent_ram_bytes(leaf,
3289 * The first search might have left us at an extent
3290 * item that ends before our target range's start, can
3291 * happen if we have holes and NO_HOLES feature enabled.
3293 if (key.offset + datal <= off) {
3296 } else if (key.offset >= off + len) {
3300 size = btrfs_item_size_nr(leaf, slot);
3301 read_extent_buffer(leaf, buf,
3302 btrfs_item_ptr_offset(leaf, slot),
3305 btrfs_release_path(path);
3306 path->leave_spinning = 0;
3308 memcpy(&new_key, &key, sizeof(new_key));
3309 new_key.objectid = btrfs_ino(inode);
3310 if (off <= key.offset)
3311 new_key.offset = key.offset + destoff - off;
3313 new_key.offset = destoff;
3316 * Deal with a hole that doesn't have an extent item
3317 * that represents it (NO_HOLES feature enabled).
3318 * This hole is either in the middle of the cloning
3319 * range or at the beginning (fully overlaps it or
3320 * partially overlaps it).
3322 if (new_key.offset != last_dest_end)
3323 drop_start = last_dest_end;
3325 drop_start = new_key.offset;
3328 * 1 - adjusting old extent (we may have to split it)
3329 * 1 - add new extent
3332 trans = btrfs_start_transaction(root, 3);
3333 if (IS_ERR(trans)) {
3334 ret = PTR_ERR(trans);
3338 if (type == BTRFS_FILE_EXTENT_REG ||
3339 type == BTRFS_FILE_EXTENT_PREALLOC) {
3341 * a | --- range to clone ---| b
3342 * | ------------- extent ------------- |
3345 /* subtract range b */
3346 if (key.offset + datal > off + len)
3347 datal = off + len - key.offset;
3349 /* subtract range a */
3350 if (off > key.offset) {
3351 datao += off - key.offset;
3352 datal -= off - key.offset;
3355 ret = btrfs_drop_extents(trans, root, inode,
3357 new_key.offset + datal,
3360 if (ret != -EOPNOTSUPP)
3361 btrfs_abort_transaction(trans,
3363 btrfs_end_transaction(trans, root);
3367 ret = btrfs_insert_empty_item(trans, root, path,
3370 btrfs_abort_transaction(trans, root,
3372 btrfs_end_transaction(trans, root);
3376 leaf = path->nodes[0];
3377 slot = path->slots[0];
3378 write_extent_buffer(leaf, buf,
3379 btrfs_item_ptr_offset(leaf, slot),
3382 extent = btrfs_item_ptr(leaf, slot,
3383 struct btrfs_file_extent_item);
3385 /* disko == 0 means it's a hole */
3389 btrfs_set_file_extent_offset(leaf, extent,
3391 btrfs_set_file_extent_num_bytes(leaf, extent,
3395 * We need to look up the roots that point at
3396 * this bytenr and see if the new root does. If
3397 * it does not we need to make sure we update
3398 * quotas appropriately.
3400 if (disko && root != BTRFS_I(src)->root &&
3401 disko != last_disko) {
3402 no_quota = check_ref(trans, root,
3405 btrfs_abort_transaction(trans,
3408 btrfs_end_transaction(trans,
3416 inode_add_bytes(inode, datal);
3417 ret = btrfs_inc_extent_ref(trans, root,
3419 root->root_key.objectid,
3421 new_key.offset - datao,
3424 btrfs_abort_transaction(trans,
3427 btrfs_end_transaction(trans,
3433 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3436 u64 aligned_end = 0;
3438 if (off > key.offset) {
3439 skip = off - key.offset;
3440 new_key.offset += skip;
3443 if (key.offset + datal > off + len)
3444 trim = key.offset + datal - (off + len);
3446 if (comp && (skip || trim)) {
3448 btrfs_end_transaction(trans, root);
3451 size -= skip + trim;
3452 datal -= skip + trim;
3454 aligned_end = ALIGN(new_key.offset + datal,
3456 ret = btrfs_drop_extents(trans, root, inode,
3461 if (ret != -EOPNOTSUPP)
3462 btrfs_abort_transaction(trans,
3464 btrfs_end_transaction(trans, root);
3468 ret = btrfs_insert_empty_item(trans, root, path,
3471 btrfs_abort_transaction(trans, root,
3473 btrfs_end_transaction(trans, root);
3479 btrfs_file_extent_calc_inline_size(0);
3480 memmove(buf+start, buf+start+skip,
3484 leaf = path->nodes[0];
3485 slot = path->slots[0];
3486 write_extent_buffer(leaf, buf,
3487 btrfs_item_ptr_offset(leaf, slot),
3489 inode_add_bytes(inode, datal);
3492 /* If we have an implicit hole (NO_HOLES feature). */
3493 if (drop_start < new_key.offset)
3494 clone_update_extent_map(inode, trans,
3496 new_key.offset - drop_start);
3498 clone_update_extent_map(inode, trans, path, 0, 0);
3500 btrfs_mark_buffer_dirty(leaf);
3501 btrfs_release_path(path);
3503 last_dest_end = ALIGN(new_key.offset + datal,
3505 ret = clone_finish_inode_update(trans, inode,
3510 if (new_key.offset + datal >= destoff + len)
3513 btrfs_release_path(path);
3518 if (last_dest_end < destoff + len) {
3520 * We have an implicit hole (NO_HOLES feature is enabled) that
3521 * fully or partially overlaps our cloning range at its end.
3523 btrfs_release_path(path);
3526 * 1 - remove extent(s)
3529 trans = btrfs_start_transaction(root, 2);
3530 if (IS_ERR(trans)) {
3531 ret = PTR_ERR(trans);
3534 ret = btrfs_drop_extents(trans, root, inode,
3535 last_dest_end, destoff + len, 1);
3537 if (ret != -EOPNOTSUPP)
3538 btrfs_abort_transaction(trans, root, ret);
3539 btrfs_end_transaction(trans, root);
3542 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3543 destoff + len - last_dest_end);
3544 ret = clone_finish_inode_update(trans, inode, destoff + len,
3549 btrfs_free_path(path);
3554 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3555 u64 off, u64 olen, u64 destoff)
3557 struct inode *inode = file_inode(file);
3558 struct btrfs_root *root = BTRFS_I(inode)->root;
3563 u64 bs = root->fs_info->sb->s_blocksize;
3568 * - split compressed inline extents. annoying: we need to
3569 * decompress into destination's address_space (the file offset
3570 * may change, so source mapping won't do), then recompress (or
3571 * otherwise reinsert) a subrange.
3573 * - split destination inode's inline extents. The inline extents can
3574 * be either compressed or non-compressed.
3577 /* the destination must be opened for writing */
3578 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3581 if (btrfs_root_readonly(root))
3584 ret = mnt_want_write_file(file);
3588 src_file = fdget(srcfd);
3589 if (!src_file.file) {
3591 goto out_drop_write;
3595 if (src_file.file->f_path.mnt != file->f_path.mnt)
3598 src = file_inode(src_file.file);
3604 /* the src must be open for reading */
3605 if (!(src_file.file->f_mode & FMODE_READ))
3608 /* don't make the dst file partly checksummed */
3609 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3610 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3614 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3618 if (src->i_sb != inode->i_sb)
3623 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3624 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3626 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3627 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3630 mutex_lock(&src->i_mutex);
3633 /* determine range to clone */
3635 if (off + len > src->i_size || off + len < off)
3638 olen = len = src->i_size - off;
3639 /* if we extend to eof, continue to block boundary */
3640 if (off + len == src->i_size)
3641 len = ALIGN(src->i_size, bs) - off;
3643 /* verify the end result is block aligned */
3644 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3645 !IS_ALIGNED(destoff, bs))
3648 /* verify if ranges are overlapped within the same file */
3650 if (destoff + len > off && destoff < off + len)
3654 if (destoff > inode->i_size) {
3655 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3661 * Lock the target range too. Right after we replace the file extent
3662 * items in the fs tree (which now point to the cloned data), we might
3663 * have a worker replace them with extent items relative to a write
3664 * operation that was issued before this clone operation (i.e. confront
3665 * with inode.c:btrfs_finish_ordered_io).
3668 u64 lock_start = min_t(u64, off, destoff);
3669 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3671 lock_extent_range(src, lock_start, lock_len);
3673 lock_extent_range(src, off, len);
3674 lock_extent_range(inode, destoff, len);
3677 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3680 u64 lock_start = min_t(u64, off, destoff);
3681 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3683 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3685 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3686 unlock_extent(&BTRFS_I(inode)->io_tree, destoff,
3690 * Truncate page cache pages so that future reads will see the cloned
3691 * data immediately and not the previous data.
3693 truncate_inode_pages_range(&inode->i_data, destoff,
3694 PAGE_CACHE_ALIGN(destoff + len) - 1);
3698 mutex_unlock(&src->i_mutex);
3699 mutex_unlock(&inode->i_mutex);
3701 mutex_unlock(&inode->i_mutex);
3702 mutex_unlock(&src->i_mutex);
3705 mutex_unlock(&src->i_mutex);
3710 mnt_drop_write_file(file);
3714 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
3716 struct btrfs_ioctl_clone_range_args args;
3718 if (copy_from_user(&args, argp, sizeof(args)))
3720 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3721 args.src_length, args.dest_offset);
3725 * there are many ways the trans_start and trans_end ioctls can lead
3726 * to deadlocks. They should only be used by applications that
3727 * basically own the machine, and have a very in depth understanding
3728 * of all the possible deadlocks and enospc problems.
3730 static long btrfs_ioctl_trans_start(struct file *file)
3732 struct inode *inode = file_inode(file);
3733 struct btrfs_root *root = BTRFS_I(inode)->root;
3734 struct btrfs_trans_handle *trans;
3738 if (!capable(CAP_SYS_ADMIN))
3742 if (file->private_data)
3746 if (btrfs_root_readonly(root))
3749 ret = mnt_want_write_file(file);
3753 atomic_inc(&root->fs_info->open_ioctl_trans);
3756 trans = btrfs_start_ioctl_transaction(root);
3760 file->private_data = trans;
3764 atomic_dec(&root->fs_info->open_ioctl_trans);
3765 mnt_drop_write_file(file);
3770 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3772 struct inode *inode = file_inode(file);
3773 struct btrfs_root *root = BTRFS_I(inode)->root;
3774 struct btrfs_root *new_root;
3775 struct btrfs_dir_item *di;
3776 struct btrfs_trans_handle *trans;
3777 struct btrfs_path *path;
3778 struct btrfs_key location;
3779 struct btrfs_disk_key disk_key;
3784 if (!capable(CAP_SYS_ADMIN))
3787 ret = mnt_want_write_file(file);
3791 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3797 objectid = BTRFS_FS_TREE_OBJECTID;
3799 location.objectid = objectid;
3800 location.type = BTRFS_ROOT_ITEM_KEY;
3801 location.offset = (u64)-1;
3803 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3804 if (IS_ERR(new_root)) {
3805 ret = PTR_ERR(new_root);
3809 path = btrfs_alloc_path();
3814 path->leave_spinning = 1;
3816 trans = btrfs_start_transaction(root, 1);
3817 if (IS_ERR(trans)) {
3818 btrfs_free_path(path);
3819 ret = PTR_ERR(trans);
3823 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
3824 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3825 dir_id, "default", 7, 1);
3826 if (IS_ERR_OR_NULL(di)) {
3827 btrfs_free_path(path);
3828 btrfs_end_transaction(trans, root);
3829 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3830 "item, this isn't going to work");
3835 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3836 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3837 btrfs_mark_buffer_dirty(path->nodes[0]);
3838 btrfs_free_path(path);
3840 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
3841 btrfs_end_transaction(trans, root);
3843 mnt_drop_write_file(file);
3847 void btrfs_get_block_group_info(struct list_head *groups_list,
3848 struct btrfs_ioctl_space_info *space)
3850 struct btrfs_block_group_cache *block_group;
3852 space->total_bytes = 0;
3853 space->used_bytes = 0;
3855 list_for_each_entry(block_group, groups_list, list) {
3856 space->flags = block_group->flags;
3857 space->total_bytes += block_group->key.offset;
3858 space->used_bytes +=
3859 btrfs_block_group_used(&block_group->item);
3863 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3865 struct btrfs_ioctl_space_args space_args;
3866 struct btrfs_ioctl_space_info space;
3867 struct btrfs_ioctl_space_info *dest;
3868 struct btrfs_ioctl_space_info *dest_orig;
3869 struct btrfs_ioctl_space_info __user *user_dest;
3870 struct btrfs_space_info *info;
3871 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3872 BTRFS_BLOCK_GROUP_SYSTEM,
3873 BTRFS_BLOCK_GROUP_METADATA,
3874 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3881 if (copy_from_user(&space_args,
3882 (struct btrfs_ioctl_space_args __user *)arg,
3883 sizeof(space_args)))
3886 for (i = 0; i < num_types; i++) {
3887 struct btrfs_space_info *tmp;
3891 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3893 if (tmp->flags == types[i]) {
3903 down_read(&info->groups_sem);
3904 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3905 if (!list_empty(&info->block_groups[c]))
3908 up_read(&info->groups_sem);
3912 * Global block reserve, exported as a space_info
3916 /* space_slots == 0 means they are asking for a count */
3917 if (space_args.space_slots == 0) {
3918 space_args.total_spaces = slot_count;
3922 slot_count = min_t(u64, space_args.space_slots, slot_count);
3924 alloc_size = sizeof(*dest) * slot_count;
3926 /* we generally have at most 6 or so space infos, one for each raid
3927 * level. So, a whole page should be more than enough for everyone
3929 if (alloc_size > PAGE_CACHE_SIZE)
3932 space_args.total_spaces = 0;
3933 dest = kmalloc(alloc_size, GFP_NOFS);
3938 /* now we have a buffer to copy into */
3939 for (i = 0; i < num_types; i++) {
3940 struct btrfs_space_info *tmp;
3947 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3949 if (tmp->flags == types[i]) {
3958 down_read(&info->groups_sem);
3959 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3960 if (!list_empty(&info->block_groups[c])) {
3961 btrfs_get_block_group_info(
3962 &info->block_groups[c], &space);
3963 memcpy(dest, &space, sizeof(space));
3965 space_args.total_spaces++;
3971 up_read(&info->groups_sem);
3975 * Add global block reserve
3978 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
3980 spin_lock(&block_rsv->lock);
3981 space.total_bytes = block_rsv->size;
3982 space.used_bytes = block_rsv->size - block_rsv->reserved;
3983 spin_unlock(&block_rsv->lock);
3984 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3985 memcpy(dest, &space, sizeof(space));
3986 space_args.total_spaces++;
3989 user_dest = (struct btrfs_ioctl_space_info __user *)
3990 (arg + sizeof(struct btrfs_ioctl_space_args));
3992 if (copy_to_user(user_dest, dest_orig, alloc_size))
3997 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4004 * there are many ways the trans_start and trans_end ioctls can lead
4005 * to deadlocks. They should only be used by applications that
4006 * basically own the machine, and have a very in depth understanding
4007 * of all the possible deadlocks and enospc problems.
4009 long btrfs_ioctl_trans_end(struct file *file)
4011 struct inode *inode = file_inode(file);
4012 struct btrfs_root *root = BTRFS_I(inode)->root;
4013 struct btrfs_trans_handle *trans;
4015 trans = file->private_data;
4018 file->private_data = NULL;
4020 btrfs_end_transaction(trans, root);
4022 atomic_dec(&root->fs_info->open_ioctl_trans);
4024 mnt_drop_write_file(file);
4028 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4031 struct btrfs_trans_handle *trans;
4035 trans = btrfs_attach_transaction_barrier(root);
4036 if (IS_ERR(trans)) {
4037 if (PTR_ERR(trans) != -ENOENT)
4038 return PTR_ERR(trans);
4040 /* No running transaction, don't bother */
4041 transid = root->fs_info->last_trans_committed;
4044 transid = trans->transid;
4045 ret = btrfs_commit_transaction_async(trans, root, 0);
4047 btrfs_end_transaction(trans, root);
4052 if (copy_to_user(argp, &transid, sizeof(transid)))
4057 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4063 if (copy_from_user(&transid, argp, sizeof(transid)))
4066 transid = 0; /* current trans */
4068 return btrfs_wait_for_commit(root, transid);
4071 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4073 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4074 struct btrfs_ioctl_scrub_args *sa;
4077 if (!capable(CAP_SYS_ADMIN))
4080 sa = memdup_user(arg, sizeof(*sa));
4084 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4085 ret = mnt_want_write_file(file);
4090 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
4091 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4094 if (copy_to_user(arg, sa, sizeof(*sa)))
4097 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4098 mnt_drop_write_file(file);
4104 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4106 if (!capable(CAP_SYS_ADMIN))
4109 return btrfs_scrub_cancel(root->fs_info);
4112 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4115 struct btrfs_ioctl_scrub_args *sa;
4118 if (!capable(CAP_SYS_ADMIN))
4121 sa = memdup_user(arg, sizeof(*sa));
4125 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4127 if (copy_to_user(arg, sa, sizeof(*sa)))
4134 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
4137 struct btrfs_ioctl_get_dev_stats *sa;
4140 sa = memdup_user(arg, sizeof(*sa));
4144 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4149 ret = btrfs_get_dev_stats(root, sa);
4151 if (copy_to_user(arg, sa, sizeof(*sa)))
4158 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4160 struct btrfs_ioctl_dev_replace_args *p;
4163 if (!capable(CAP_SYS_ADMIN))
4166 p = memdup_user(arg, sizeof(*p));
4171 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4172 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4177 &root->fs_info->mutually_exclusive_operation_running,
4179 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4181 ret = btrfs_dev_replace_start(root, p);
4183 &root->fs_info->mutually_exclusive_operation_running,
4187 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4188 btrfs_dev_replace_status(root->fs_info, p);
4191 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4192 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4199 if (copy_to_user(arg, p, sizeof(*p)))
4206 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4212 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4213 struct inode_fs_paths *ipath = NULL;
4214 struct btrfs_path *path;
4216 if (!capable(CAP_DAC_READ_SEARCH))
4219 path = btrfs_alloc_path();
4225 ipa = memdup_user(arg, sizeof(*ipa));
4232 size = min_t(u32, ipa->size, 4096);
4233 ipath = init_ipath(size, root, path);
4234 if (IS_ERR(ipath)) {
4235 ret = PTR_ERR(ipath);
4240 ret = paths_from_inode(ipa->inum, ipath);
4244 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4245 rel_ptr = ipath->fspath->val[i] -
4246 (u64)(unsigned long)ipath->fspath->val;
4247 ipath->fspath->val[i] = rel_ptr;
4250 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4251 (void *)(unsigned long)ipath->fspath, size);
4258 btrfs_free_path(path);
4265 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4267 struct btrfs_data_container *inodes = ctx;
4268 const size_t c = 3 * sizeof(u64);
4270 if (inodes->bytes_left >= c) {
4271 inodes->bytes_left -= c;
4272 inodes->val[inodes->elem_cnt] = inum;
4273 inodes->val[inodes->elem_cnt + 1] = offset;
4274 inodes->val[inodes->elem_cnt + 2] = root;
4275 inodes->elem_cnt += 3;
4277 inodes->bytes_missing += c - inodes->bytes_left;
4278 inodes->bytes_left = 0;
4279 inodes->elem_missed += 3;
4285 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4290 struct btrfs_ioctl_logical_ino_args *loi;
4291 struct btrfs_data_container *inodes = NULL;
4292 struct btrfs_path *path = NULL;
4294 if (!capable(CAP_SYS_ADMIN))
4297 loi = memdup_user(arg, sizeof(*loi));
4304 path = btrfs_alloc_path();
4310 size = min_t(u32, loi->size, 64 * 1024);
4311 inodes = init_data_container(size);
4312 if (IS_ERR(inodes)) {
4313 ret = PTR_ERR(inodes);
4318 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4319 build_ino_list, inodes);
4325 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4326 (void *)(unsigned long)inodes, size);
4331 btrfs_free_path(path);
4338 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4339 struct btrfs_ioctl_balance_args *bargs)
4341 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4343 bargs->flags = bctl->flags;
4345 if (atomic_read(&fs_info->balance_running))
4346 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4347 if (atomic_read(&fs_info->balance_pause_req))
4348 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4349 if (atomic_read(&fs_info->balance_cancel_req))
4350 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4352 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4353 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4354 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4357 spin_lock(&fs_info->balance_lock);
4358 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4359 spin_unlock(&fs_info->balance_lock);
4361 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4365 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4367 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4368 struct btrfs_fs_info *fs_info = root->fs_info;
4369 struct btrfs_ioctl_balance_args *bargs;
4370 struct btrfs_balance_control *bctl;
4371 bool need_unlock; /* for mut. excl. ops lock */
4374 if (!capable(CAP_SYS_ADMIN))
4377 ret = mnt_want_write_file(file);
4382 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4383 mutex_lock(&fs_info->volume_mutex);
4384 mutex_lock(&fs_info->balance_mutex);
4390 * mut. excl. ops lock is locked. Three possibilites:
4391 * (1) some other op is running
4392 * (2) balance is running
4393 * (3) balance is paused -- special case (think resume)
4395 mutex_lock(&fs_info->balance_mutex);
4396 if (fs_info->balance_ctl) {
4397 /* this is either (2) or (3) */
4398 if (!atomic_read(&fs_info->balance_running)) {
4399 mutex_unlock(&fs_info->balance_mutex);
4400 if (!mutex_trylock(&fs_info->volume_mutex))
4402 mutex_lock(&fs_info->balance_mutex);
4404 if (fs_info->balance_ctl &&
4405 !atomic_read(&fs_info->balance_running)) {
4407 need_unlock = false;
4411 mutex_unlock(&fs_info->balance_mutex);
4412 mutex_unlock(&fs_info->volume_mutex);
4416 mutex_unlock(&fs_info->balance_mutex);
4422 mutex_unlock(&fs_info->balance_mutex);
4423 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4428 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4431 bargs = memdup_user(arg, sizeof(*bargs));
4432 if (IS_ERR(bargs)) {
4433 ret = PTR_ERR(bargs);
4437 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4438 if (!fs_info->balance_ctl) {
4443 bctl = fs_info->balance_ctl;
4444 spin_lock(&fs_info->balance_lock);
4445 bctl->flags |= BTRFS_BALANCE_RESUME;
4446 spin_unlock(&fs_info->balance_lock);
4454 if (fs_info->balance_ctl) {
4459 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4465 bctl->fs_info = fs_info;
4467 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4468 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4469 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4471 bctl->flags = bargs->flags;
4473 /* balance everything - no filters */
4474 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4479 * Ownership of bctl and mutually_exclusive_operation_running
4480 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4481 * or, if restriper was paused all the way until unmount, in
4482 * free_fs_info. mutually_exclusive_operation_running is
4483 * cleared in __cancel_balance.
4485 need_unlock = false;
4487 ret = btrfs_balance(bctl, bargs);
4490 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4497 mutex_unlock(&fs_info->balance_mutex);
4498 mutex_unlock(&fs_info->volume_mutex);
4500 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4502 mnt_drop_write_file(file);
4506 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4508 if (!capable(CAP_SYS_ADMIN))
4512 case BTRFS_BALANCE_CTL_PAUSE:
4513 return btrfs_pause_balance(root->fs_info);
4514 case BTRFS_BALANCE_CTL_CANCEL:
4515 return btrfs_cancel_balance(root->fs_info);
4521 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4524 struct btrfs_fs_info *fs_info = root->fs_info;
4525 struct btrfs_ioctl_balance_args *bargs;
4528 if (!capable(CAP_SYS_ADMIN))
4531 mutex_lock(&fs_info->balance_mutex);
4532 if (!fs_info->balance_ctl) {
4537 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4543 update_ioctl_balance_args(fs_info, 1, bargs);
4545 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4550 mutex_unlock(&fs_info->balance_mutex);
4554 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4556 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4557 struct btrfs_ioctl_quota_ctl_args *sa;
4558 struct btrfs_trans_handle *trans = NULL;
4562 if (!capable(CAP_SYS_ADMIN))
4565 ret = mnt_want_write_file(file);
4569 sa = memdup_user(arg, sizeof(*sa));
4575 down_write(&root->fs_info->subvol_sem);
4576 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4577 if (IS_ERR(trans)) {
4578 ret = PTR_ERR(trans);
4583 case BTRFS_QUOTA_CTL_ENABLE:
4584 ret = btrfs_quota_enable(trans, root->fs_info);
4586 case BTRFS_QUOTA_CTL_DISABLE:
4587 ret = btrfs_quota_disable(trans, root->fs_info);
4594 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4599 up_write(&root->fs_info->subvol_sem);
4601 mnt_drop_write_file(file);
4605 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4607 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4608 struct btrfs_ioctl_qgroup_assign_args *sa;
4609 struct btrfs_trans_handle *trans;
4613 if (!capable(CAP_SYS_ADMIN))
4616 ret = mnt_want_write_file(file);
4620 sa = memdup_user(arg, sizeof(*sa));
4626 trans = btrfs_join_transaction(root);
4627 if (IS_ERR(trans)) {
4628 ret = PTR_ERR(trans);
4632 /* FIXME: check if the IDs really exist */
4634 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4637 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4641 err = btrfs_end_transaction(trans, root);
4648 mnt_drop_write_file(file);
4652 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4654 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4655 struct btrfs_ioctl_qgroup_create_args *sa;
4656 struct btrfs_trans_handle *trans;
4660 if (!capable(CAP_SYS_ADMIN))
4663 ret = mnt_want_write_file(file);
4667 sa = memdup_user(arg, sizeof(*sa));
4673 if (!sa->qgroupid) {
4678 trans = btrfs_join_transaction(root);
4679 if (IS_ERR(trans)) {
4680 ret = PTR_ERR(trans);
4684 /* FIXME: check if the IDs really exist */
4686 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4689 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4692 err = btrfs_end_transaction(trans, root);
4699 mnt_drop_write_file(file);
4703 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4705 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4706 struct btrfs_ioctl_qgroup_limit_args *sa;
4707 struct btrfs_trans_handle *trans;
4712 if (!capable(CAP_SYS_ADMIN))
4715 ret = mnt_want_write_file(file);
4719 sa = memdup_user(arg, sizeof(*sa));
4725 trans = btrfs_join_transaction(root);
4726 if (IS_ERR(trans)) {
4727 ret = PTR_ERR(trans);
4731 qgroupid = sa->qgroupid;
4733 /* take the current subvol as qgroup */
4734 qgroupid = root->root_key.objectid;
4737 /* FIXME: check if the IDs really exist */
4738 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4740 err = btrfs_end_transaction(trans, root);
4747 mnt_drop_write_file(file);
4751 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4753 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4754 struct btrfs_ioctl_quota_rescan_args *qsa;
4757 if (!capable(CAP_SYS_ADMIN))
4760 ret = mnt_want_write_file(file);
4764 qsa = memdup_user(arg, sizeof(*qsa));
4775 ret = btrfs_qgroup_rescan(root->fs_info);
4780 mnt_drop_write_file(file);
4784 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4786 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4787 struct btrfs_ioctl_quota_rescan_args *qsa;
4790 if (!capable(CAP_SYS_ADMIN))
4793 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4797 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4799 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4802 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4809 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4811 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4813 if (!capable(CAP_SYS_ADMIN))
4816 return btrfs_qgroup_wait_for_completion(root->fs_info);
4819 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4820 struct btrfs_ioctl_received_subvol_args *sa)
4822 struct inode *inode = file_inode(file);
4823 struct btrfs_root *root = BTRFS_I(inode)->root;
4824 struct btrfs_root_item *root_item = &root->root_item;
4825 struct btrfs_trans_handle *trans;
4826 struct timespec ct = CURRENT_TIME;
4828 int received_uuid_changed;
4830 if (!inode_owner_or_capable(inode))
4833 ret = mnt_want_write_file(file);
4837 down_write(&root->fs_info->subvol_sem);
4839 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4844 if (btrfs_root_readonly(root)) {
4851 * 2 - uuid items (received uuid + subvol uuid)
4853 trans = btrfs_start_transaction(root, 3);
4854 if (IS_ERR(trans)) {
4855 ret = PTR_ERR(trans);
4860 sa->rtransid = trans->transid;
4861 sa->rtime.sec = ct.tv_sec;
4862 sa->rtime.nsec = ct.tv_nsec;
4864 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4866 if (received_uuid_changed &&
4867 !btrfs_is_empty_uuid(root_item->received_uuid))
4868 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4869 root_item->received_uuid,
4870 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4871 root->root_key.objectid);
4872 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4873 btrfs_set_root_stransid(root_item, sa->stransid);
4874 btrfs_set_root_rtransid(root_item, sa->rtransid);
4875 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4876 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4877 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4878 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4880 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4881 &root->root_key, &root->root_item);
4883 btrfs_end_transaction(trans, root);
4886 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4887 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4889 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4890 root->root_key.objectid);
4891 if (ret < 0 && ret != -EEXIST) {
4892 btrfs_abort_transaction(trans, root, ret);
4896 ret = btrfs_commit_transaction(trans, root);
4898 btrfs_abort_transaction(trans, root, ret);
4903 up_write(&root->fs_info->subvol_sem);
4904 mnt_drop_write_file(file);
4909 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4912 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4913 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4916 args32 = memdup_user(arg, sizeof(*args32));
4917 if (IS_ERR(args32)) {
4918 ret = PTR_ERR(args32);
4923 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
4929 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4930 args64->stransid = args32->stransid;
4931 args64->rtransid = args32->rtransid;
4932 args64->stime.sec = args32->stime.sec;
4933 args64->stime.nsec = args32->stime.nsec;
4934 args64->rtime.sec = args32->rtime.sec;
4935 args64->rtime.nsec = args32->rtime.nsec;
4936 args64->flags = args32->flags;
4938 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4942 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4943 args32->stransid = args64->stransid;
4944 args32->rtransid = args64->rtransid;
4945 args32->stime.sec = args64->stime.sec;
4946 args32->stime.nsec = args64->stime.nsec;
4947 args32->rtime.sec = args64->rtime.sec;
4948 args32->rtime.nsec = args64->rtime.nsec;
4949 args32->flags = args64->flags;
4951 ret = copy_to_user(arg, args32, sizeof(*args32));
4962 static long btrfs_ioctl_set_received_subvol(struct file *file,
4965 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4968 sa = memdup_user(arg, sizeof(*sa));
4975 ret = _btrfs_ioctl_set_received_subvol(file, sa);
4980 ret = copy_to_user(arg, sa, sizeof(*sa));
4989 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4991 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4994 char label[BTRFS_LABEL_SIZE];
4996 spin_lock(&root->fs_info->super_lock);
4997 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4998 spin_unlock(&root->fs_info->super_lock);
5000 len = strnlen(label, BTRFS_LABEL_SIZE);
5002 if (len == BTRFS_LABEL_SIZE) {
5003 btrfs_warn(root->fs_info,
5004 "label is too long, return the first %zu bytes", --len);
5007 ret = copy_to_user(arg, label, len);
5009 return ret ? -EFAULT : 0;
5012 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5014 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5015 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5016 struct btrfs_trans_handle *trans;
5017 char label[BTRFS_LABEL_SIZE];
5020 if (!capable(CAP_SYS_ADMIN))
5023 if (copy_from_user(label, arg, sizeof(label)))
5026 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5027 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
5028 BTRFS_LABEL_SIZE - 1);
5032 ret = mnt_want_write_file(file);
5036 trans = btrfs_start_transaction(root, 0);
5037 if (IS_ERR(trans)) {
5038 ret = PTR_ERR(trans);
5042 spin_lock(&root->fs_info->super_lock);
5043 strcpy(super_block->label, label);
5044 spin_unlock(&root->fs_info->super_lock);
5045 ret = btrfs_commit_transaction(trans, root);
5048 mnt_drop_write_file(file);
5052 #define INIT_FEATURE_FLAGS(suffix) \
5053 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5054 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5055 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5057 static int btrfs_ioctl_get_supported_features(struct file *file,
5060 static struct btrfs_ioctl_feature_flags features[3] = {
5061 INIT_FEATURE_FLAGS(SUPP),
5062 INIT_FEATURE_FLAGS(SAFE_SET),
5063 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5066 if (copy_to_user(arg, &features, sizeof(features)))
5072 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5074 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5075 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5076 struct btrfs_ioctl_feature_flags features;
5078 features.compat_flags = btrfs_super_compat_flags(super_block);
5079 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5080 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5082 if (copy_to_user(arg, &features, sizeof(features)))
5088 static int check_feature_bits(struct btrfs_root *root,
5089 enum btrfs_feature_set set,
5090 u64 change_mask, u64 flags, u64 supported_flags,
5091 u64 safe_set, u64 safe_clear)
5093 const char *type = btrfs_feature_set_names[set];
5095 u64 disallowed, unsupported;
5096 u64 set_mask = flags & change_mask;
5097 u64 clear_mask = ~flags & change_mask;
5099 unsupported = set_mask & ~supported_flags;
5101 names = btrfs_printable_features(set, unsupported);
5103 btrfs_warn(root->fs_info,
5104 "this kernel does not support the %s feature bit%s",
5105 names, strchr(names, ',') ? "s" : "");
5108 btrfs_warn(root->fs_info,
5109 "this kernel does not support %s bits 0x%llx",
5114 disallowed = set_mask & ~safe_set;
5116 names = btrfs_printable_features(set, disallowed);
5118 btrfs_warn(root->fs_info,
5119 "can't set the %s feature bit%s while mounted",
5120 names, strchr(names, ',') ? "s" : "");
5123 btrfs_warn(root->fs_info,
5124 "can't set %s bits 0x%llx while mounted",
5129 disallowed = clear_mask & ~safe_clear;
5131 names = btrfs_printable_features(set, disallowed);
5133 btrfs_warn(root->fs_info,
5134 "can't clear the %s feature bit%s while mounted",
5135 names, strchr(names, ',') ? "s" : "");
5138 btrfs_warn(root->fs_info,
5139 "can't clear %s bits 0x%llx while mounted",
5147 #define check_feature(root, change_mask, flags, mask_base) \
5148 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5149 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5150 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5151 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5153 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5155 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5156 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5157 struct btrfs_ioctl_feature_flags flags[2];
5158 struct btrfs_trans_handle *trans;
5162 if (!capable(CAP_SYS_ADMIN))
5165 if (copy_from_user(flags, arg, sizeof(flags)))
5169 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5170 !flags[0].incompat_flags)
5173 ret = check_feature(root, flags[0].compat_flags,
5174 flags[1].compat_flags, COMPAT);
5178 ret = check_feature(root, flags[0].compat_ro_flags,
5179 flags[1].compat_ro_flags, COMPAT_RO);
5183 ret = check_feature(root, flags[0].incompat_flags,
5184 flags[1].incompat_flags, INCOMPAT);
5188 trans = btrfs_start_transaction(root, 0);
5190 return PTR_ERR(trans);
5192 spin_lock(&root->fs_info->super_lock);
5193 newflags = btrfs_super_compat_flags(super_block);
5194 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5195 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5196 btrfs_set_super_compat_flags(super_block, newflags);
5198 newflags = btrfs_super_compat_ro_flags(super_block);
5199 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5200 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5201 btrfs_set_super_compat_ro_flags(super_block, newflags);
5203 newflags = btrfs_super_incompat_flags(super_block);
5204 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5205 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5206 btrfs_set_super_incompat_flags(super_block, newflags);
5207 spin_unlock(&root->fs_info->super_lock);
5209 return btrfs_commit_transaction(trans, root);
5212 long btrfs_ioctl(struct file *file, unsigned int
5213 cmd, unsigned long arg)
5215 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5216 void __user *argp = (void __user *)arg;
5219 case FS_IOC_GETFLAGS:
5220 return btrfs_ioctl_getflags(file, argp);
5221 case FS_IOC_SETFLAGS:
5222 return btrfs_ioctl_setflags(file, argp);
5223 case FS_IOC_GETVERSION:
5224 return btrfs_ioctl_getversion(file, argp);
5226 return btrfs_ioctl_fitrim(file, argp);
5227 case BTRFS_IOC_SNAP_CREATE:
5228 return btrfs_ioctl_snap_create(file, argp, 0);
5229 case BTRFS_IOC_SNAP_CREATE_V2:
5230 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5231 case BTRFS_IOC_SUBVOL_CREATE:
5232 return btrfs_ioctl_snap_create(file, argp, 1);
5233 case BTRFS_IOC_SUBVOL_CREATE_V2:
5234 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5235 case BTRFS_IOC_SNAP_DESTROY:
5236 return btrfs_ioctl_snap_destroy(file, argp);
5237 case BTRFS_IOC_SUBVOL_GETFLAGS:
5238 return btrfs_ioctl_subvol_getflags(file, argp);
5239 case BTRFS_IOC_SUBVOL_SETFLAGS:
5240 return btrfs_ioctl_subvol_setflags(file, argp);
5241 case BTRFS_IOC_DEFAULT_SUBVOL:
5242 return btrfs_ioctl_default_subvol(file, argp);
5243 case BTRFS_IOC_DEFRAG:
5244 return btrfs_ioctl_defrag(file, NULL);
5245 case BTRFS_IOC_DEFRAG_RANGE:
5246 return btrfs_ioctl_defrag(file, argp);
5247 case BTRFS_IOC_RESIZE:
5248 return btrfs_ioctl_resize(file, argp);
5249 case BTRFS_IOC_ADD_DEV:
5250 return btrfs_ioctl_add_dev(root, argp);
5251 case BTRFS_IOC_RM_DEV:
5252 return btrfs_ioctl_rm_dev(file, argp);
5253 case BTRFS_IOC_FS_INFO:
5254 return btrfs_ioctl_fs_info(root, argp);
5255 case BTRFS_IOC_DEV_INFO:
5256 return btrfs_ioctl_dev_info(root, argp);
5257 case BTRFS_IOC_BALANCE:
5258 return btrfs_ioctl_balance(file, NULL);
5259 case BTRFS_IOC_CLONE:
5260 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
5261 case BTRFS_IOC_CLONE_RANGE:
5262 return btrfs_ioctl_clone_range(file, argp);
5263 case BTRFS_IOC_TRANS_START:
5264 return btrfs_ioctl_trans_start(file);
5265 case BTRFS_IOC_TRANS_END:
5266 return btrfs_ioctl_trans_end(file);
5267 case BTRFS_IOC_TREE_SEARCH:
5268 return btrfs_ioctl_tree_search(file, argp);
5269 case BTRFS_IOC_TREE_SEARCH_V2:
5270 return btrfs_ioctl_tree_search_v2(file, argp);
5271 case BTRFS_IOC_INO_LOOKUP:
5272 return btrfs_ioctl_ino_lookup(file, argp);
5273 case BTRFS_IOC_INO_PATHS:
5274 return btrfs_ioctl_ino_to_path(root, argp);
5275 case BTRFS_IOC_LOGICAL_INO:
5276 return btrfs_ioctl_logical_to_ino(root, argp);
5277 case BTRFS_IOC_SPACE_INFO:
5278 return btrfs_ioctl_space_info(root, argp);
5279 case BTRFS_IOC_SYNC: {
5282 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
5285 ret = btrfs_sync_fs(file->f_dentry->d_sb, 1);
5288 case BTRFS_IOC_START_SYNC:
5289 return btrfs_ioctl_start_sync(root, argp);
5290 case BTRFS_IOC_WAIT_SYNC:
5291 return btrfs_ioctl_wait_sync(root, argp);
5292 case BTRFS_IOC_SCRUB:
5293 return btrfs_ioctl_scrub(file, argp);
5294 case BTRFS_IOC_SCRUB_CANCEL:
5295 return btrfs_ioctl_scrub_cancel(root, argp);
5296 case BTRFS_IOC_SCRUB_PROGRESS:
5297 return btrfs_ioctl_scrub_progress(root, argp);
5298 case BTRFS_IOC_BALANCE_V2:
5299 return btrfs_ioctl_balance(file, argp);
5300 case BTRFS_IOC_BALANCE_CTL:
5301 return btrfs_ioctl_balance_ctl(root, arg);
5302 case BTRFS_IOC_BALANCE_PROGRESS:
5303 return btrfs_ioctl_balance_progress(root, argp);
5304 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5305 return btrfs_ioctl_set_received_subvol(file, argp);
5307 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5308 return btrfs_ioctl_set_received_subvol_32(file, argp);
5310 case BTRFS_IOC_SEND:
5311 return btrfs_ioctl_send(file, argp);
5312 case BTRFS_IOC_GET_DEV_STATS:
5313 return btrfs_ioctl_get_dev_stats(root, argp);
5314 case BTRFS_IOC_QUOTA_CTL:
5315 return btrfs_ioctl_quota_ctl(file, argp);
5316 case BTRFS_IOC_QGROUP_ASSIGN:
5317 return btrfs_ioctl_qgroup_assign(file, argp);
5318 case BTRFS_IOC_QGROUP_CREATE:
5319 return btrfs_ioctl_qgroup_create(file, argp);
5320 case BTRFS_IOC_QGROUP_LIMIT:
5321 return btrfs_ioctl_qgroup_limit(file, argp);
5322 case BTRFS_IOC_QUOTA_RESCAN:
5323 return btrfs_ioctl_quota_rescan(file, argp);
5324 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5325 return btrfs_ioctl_quota_rescan_status(file, argp);
5326 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5327 return btrfs_ioctl_quota_rescan_wait(file, argp);
5328 case BTRFS_IOC_DEV_REPLACE:
5329 return btrfs_ioctl_dev_replace(root, argp);
5330 case BTRFS_IOC_GET_FSLABEL:
5331 return btrfs_ioctl_get_fslabel(file, argp);
5332 case BTRFS_IOC_SET_FSLABEL:
5333 return btrfs_ioctl_set_fslabel(file, argp);
5334 case BTRFS_IOC_FILE_EXTENT_SAME:
5335 return btrfs_ioctl_file_extent_same(file, argp);
5336 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5337 return btrfs_ioctl_get_supported_features(file, argp);
5338 case BTRFS_IOC_GET_FEATURES:
5339 return btrfs_ioctl_get_features(file, argp);
5340 case BTRFS_IOC_SET_FEATURES:
5341 return btrfs_ioctl_set_features(file, argp);